linux/fs/f2fs/super.c
<<
>>
Prefs
   1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 * fs/f2fs/super.c
   4 *
   5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
   6 *             http://www.samsung.com/
   7 */
   8#include <linux/module.h>
   9#include <linux/init.h>
  10#include <linux/fs.h>
  11#include <linux/statfs.h>
  12#include <linux/buffer_head.h>
  13#include <linux/backing-dev.h>
  14#include <linux/kthread.h>
  15#include <linux/parser.h>
  16#include <linux/mount.h>
  17#include <linux/seq_file.h>
  18#include <linux/proc_fs.h>
  19#include <linux/random.h>
  20#include <linux/exportfs.h>
  21#include <linux/blkdev.h>
  22#include <linux/quotaops.h>
  23#include <linux/f2fs_fs.h>
  24#include <linux/sysfs.h>
  25#include <linux/quota.h>
  26#include <linux/unicode.h>
  27#include <linux/part_stat.h>
  28
  29#include "f2fs.h"
  30#include "node.h"
  31#include "segment.h"
  32#include "xattr.h"
  33#include "gc.h"
  34#include "trace.h"
  35
  36#define CREATE_TRACE_POINTS
  37#include <trace/events/f2fs.h>
  38
  39static struct kmem_cache *f2fs_inode_cachep;
  40
  41#ifdef CONFIG_F2FS_FAULT_INJECTION
  42
  43const char *f2fs_fault_name[FAULT_MAX] = {
  44        [FAULT_KMALLOC]         = "kmalloc",
  45        [FAULT_KVMALLOC]        = "kvmalloc",
  46        [FAULT_PAGE_ALLOC]      = "page alloc",
  47        [FAULT_PAGE_GET]        = "page get",
  48        [FAULT_ALLOC_BIO]       = "alloc bio",
  49        [FAULT_ALLOC_NID]       = "alloc nid",
  50        [FAULT_ORPHAN]          = "orphan",
  51        [FAULT_BLOCK]           = "no more block",
  52        [FAULT_DIR_DEPTH]       = "too big dir depth",
  53        [FAULT_EVICT_INODE]     = "evict_inode fail",
  54        [FAULT_TRUNCATE]        = "truncate fail",
  55        [FAULT_READ_IO]         = "read IO error",
  56        [FAULT_CHECKPOINT]      = "checkpoint error",
  57        [FAULT_DISCARD]         = "discard error",
  58        [FAULT_WRITE_IO]        = "write IO error",
  59};
  60
  61void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
  62                                                        unsigned int type)
  63{
  64        struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
  65
  66        if (rate) {
  67                atomic_set(&ffi->inject_ops, 0);
  68                ffi->inject_rate = rate;
  69        }
  70
  71        if (type)
  72                ffi->inject_type = type;
  73
  74        if (!rate && !type)
  75                memset(ffi, 0, sizeof(struct f2fs_fault_info));
  76}
  77#endif
  78
  79/* f2fs-wide shrinker description */
  80static struct shrinker f2fs_shrinker_info = {
  81        .scan_objects = f2fs_shrink_scan,
  82        .count_objects = f2fs_shrink_count,
  83        .seeks = DEFAULT_SEEKS,
  84};
  85
  86enum {
  87        Opt_gc_background,
  88        Opt_disable_roll_forward,
  89        Opt_norecovery,
  90        Opt_discard,
  91        Opt_nodiscard,
  92        Opt_noheap,
  93        Opt_heap,
  94        Opt_user_xattr,
  95        Opt_nouser_xattr,
  96        Opt_acl,
  97        Opt_noacl,
  98        Opt_active_logs,
  99        Opt_disable_ext_identify,
 100        Opt_inline_xattr,
 101        Opt_noinline_xattr,
 102        Opt_inline_xattr_size,
 103        Opt_inline_data,
 104        Opt_inline_dentry,
 105        Opt_noinline_dentry,
 106        Opt_flush_merge,
 107        Opt_noflush_merge,
 108        Opt_nobarrier,
 109        Opt_fastboot,
 110        Opt_extent_cache,
 111        Opt_noextent_cache,
 112        Opt_noinline_data,
 113        Opt_data_flush,
 114        Opt_reserve_root,
 115        Opt_resgid,
 116        Opt_resuid,
 117        Opt_mode,
 118        Opt_io_size_bits,
 119        Opt_fault_injection,
 120        Opt_fault_type,
 121        Opt_lazytime,
 122        Opt_nolazytime,
 123        Opt_quota,
 124        Opt_noquota,
 125        Opt_usrquota,
 126        Opt_grpquota,
 127        Opt_prjquota,
 128        Opt_usrjquota,
 129        Opt_grpjquota,
 130        Opt_prjjquota,
 131        Opt_offusrjquota,
 132        Opt_offgrpjquota,
 133        Opt_offprjjquota,
 134        Opt_jqfmt_vfsold,
 135        Opt_jqfmt_vfsv0,
 136        Opt_jqfmt_vfsv1,
 137        Opt_whint,
 138        Opt_alloc,
 139        Opt_fsync,
 140        Opt_test_dummy_encryption,
 141        Opt_checkpoint_disable,
 142        Opt_checkpoint_disable_cap,
 143        Opt_checkpoint_disable_cap_perc,
 144        Opt_checkpoint_enable,
 145        Opt_compress_algorithm,
 146        Opt_compress_log_size,
 147        Opt_compress_extension,
 148        Opt_err,
 149};
 150
 151static match_table_t f2fs_tokens = {
 152        {Opt_gc_background, "background_gc=%s"},
 153        {Opt_disable_roll_forward, "disable_roll_forward"},
 154        {Opt_norecovery, "norecovery"},
 155        {Opt_discard, "discard"},
 156        {Opt_nodiscard, "nodiscard"},
 157        {Opt_noheap, "no_heap"},
 158        {Opt_heap, "heap"},
 159        {Opt_user_xattr, "user_xattr"},
 160        {Opt_nouser_xattr, "nouser_xattr"},
 161        {Opt_acl, "acl"},
 162        {Opt_noacl, "noacl"},
 163        {Opt_active_logs, "active_logs=%u"},
 164        {Opt_disable_ext_identify, "disable_ext_identify"},
 165        {Opt_inline_xattr, "inline_xattr"},
 166        {Opt_noinline_xattr, "noinline_xattr"},
 167        {Opt_inline_xattr_size, "inline_xattr_size=%u"},
 168        {Opt_inline_data, "inline_data"},
 169        {Opt_inline_dentry, "inline_dentry"},
 170        {Opt_noinline_dentry, "noinline_dentry"},
 171        {Opt_flush_merge, "flush_merge"},
 172        {Opt_noflush_merge, "noflush_merge"},
 173        {Opt_nobarrier, "nobarrier"},
 174        {Opt_fastboot, "fastboot"},
 175        {Opt_extent_cache, "extent_cache"},
 176        {Opt_noextent_cache, "noextent_cache"},
 177        {Opt_noinline_data, "noinline_data"},
 178        {Opt_data_flush, "data_flush"},
 179        {Opt_reserve_root, "reserve_root=%u"},
 180        {Opt_resgid, "resgid=%u"},
 181        {Opt_resuid, "resuid=%u"},
 182        {Opt_mode, "mode=%s"},
 183        {Opt_io_size_bits, "io_bits=%u"},
 184        {Opt_fault_injection, "fault_injection=%u"},
 185        {Opt_fault_type, "fault_type=%u"},
 186        {Opt_lazytime, "lazytime"},
 187        {Opt_nolazytime, "nolazytime"},
 188        {Opt_quota, "quota"},
 189        {Opt_noquota, "noquota"},
 190        {Opt_usrquota, "usrquota"},
 191        {Opt_grpquota, "grpquota"},
 192        {Opt_prjquota, "prjquota"},
 193        {Opt_usrjquota, "usrjquota=%s"},
 194        {Opt_grpjquota, "grpjquota=%s"},
 195        {Opt_prjjquota, "prjjquota=%s"},
 196        {Opt_offusrjquota, "usrjquota="},
 197        {Opt_offgrpjquota, "grpjquota="},
 198        {Opt_offprjjquota, "prjjquota="},
 199        {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
 200        {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
 201        {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
 202        {Opt_whint, "whint_mode=%s"},
 203        {Opt_alloc, "alloc_mode=%s"},
 204        {Opt_fsync, "fsync_mode=%s"},
 205        {Opt_test_dummy_encryption, "test_dummy_encryption"},
 206        {Opt_checkpoint_disable, "checkpoint=disable"},
 207        {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
 208        {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
 209        {Opt_checkpoint_enable, "checkpoint=enable"},
 210        {Opt_compress_algorithm, "compress_algorithm=%s"},
 211        {Opt_compress_log_size, "compress_log_size=%u"},
 212        {Opt_compress_extension, "compress_extension=%s"},
 213        {Opt_err, NULL},
 214};
 215
 216void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
 217{
 218        struct va_format vaf;
 219        va_list args;
 220        int level;
 221
 222        va_start(args, fmt);
 223
 224        level = printk_get_level(fmt);
 225        vaf.fmt = printk_skip_level(fmt);
 226        vaf.va = &args;
 227        printk("%c%cF2FS-fs (%s): %pV\n",
 228               KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
 229
 230        va_end(args);
 231}
 232
 233#ifdef CONFIG_UNICODE
 234static const struct f2fs_sb_encodings {
 235        __u16 magic;
 236        char *name;
 237        char *version;
 238} f2fs_sb_encoding_map[] = {
 239        {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
 240};
 241
 242static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
 243                                 const struct f2fs_sb_encodings **encoding,
 244                                 __u16 *flags)
 245{
 246        __u16 magic = le16_to_cpu(sb->s_encoding);
 247        int i;
 248
 249        for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
 250                if (magic == f2fs_sb_encoding_map[i].magic)
 251                        break;
 252
 253        if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
 254                return -EINVAL;
 255
 256        *encoding = &f2fs_sb_encoding_map[i];
 257        *flags = le16_to_cpu(sb->s_encoding_flags);
 258
 259        return 0;
 260}
 261#endif
 262
 263static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
 264{
 265        block_t limit = min((sbi->user_block_count << 1) / 1000,
 266                        sbi->user_block_count - sbi->reserved_blocks);
 267
 268        /* limit is 0.2% */
 269        if (test_opt(sbi, RESERVE_ROOT) &&
 270                        F2FS_OPTION(sbi).root_reserved_blocks > limit) {
 271                F2FS_OPTION(sbi).root_reserved_blocks = limit;
 272                f2fs_info(sbi, "Reduce reserved blocks for root = %u",
 273                          F2FS_OPTION(sbi).root_reserved_blocks);
 274        }
 275        if (!test_opt(sbi, RESERVE_ROOT) &&
 276                (!uid_eq(F2FS_OPTION(sbi).s_resuid,
 277                                make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
 278                !gid_eq(F2FS_OPTION(sbi).s_resgid,
 279                                make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
 280                f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
 281                          from_kuid_munged(&init_user_ns,
 282                                           F2FS_OPTION(sbi).s_resuid),
 283                          from_kgid_munged(&init_user_ns,
 284                                           F2FS_OPTION(sbi).s_resgid));
 285}
 286
 287static void init_once(void *foo)
 288{
 289        struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
 290
 291        inode_init_once(&fi->vfs_inode);
 292}
 293
 294#ifdef CONFIG_QUOTA
 295static const char * const quotatypes[] = INITQFNAMES;
 296#define QTYPE2NAME(t) (quotatypes[t])
 297static int f2fs_set_qf_name(struct super_block *sb, int qtype,
 298                                                        substring_t *args)
 299{
 300        struct f2fs_sb_info *sbi = F2FS_SB(sb);
 301        char *qname;
 302        int ret = -EINVAL;
 303
 304        if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
 305                f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 306                return -EINVAL;
 307        }
 308        if (f2fs_sb_has_quota_ino(sbi)) {
 309                f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
 310                return 0;
 311        }
 312
 313        qname = match_strdup(args);
 314        if (!qname) {
 315                f2fs_err(sbi, "Not enough memory for storing quotafile name");
 316                return -ENOMEM;
 317        }
 318        if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
 319                if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
 320                        ret = 0;
 321                else
 322                        f2fs_err(sbi, "%s quota file already specified",
 323                                 QTYPE2NAME(qtype));
 324                goto errout;
 325        }
 326        if (strchr(qname, '/')) {
 327                f2fs_err(sbi, "quotafile must be on filesystem root");
 328                goto errout;
 329        }
 330        F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
 331        set_opt(sbi, QUOTA);
 332        return 0;
 333errout:
 334        kvfree(qname);
 335        return ret;
 336}
 337
 338static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
 339{
 340        struct f2fs_sb_info *sbi = F2FS_SB(sb);
 341
 342        if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
 343                f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
 344                return -EINVAL;
 345        }
 346        kvfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
 347        F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
 348        return 0;
 349}
 350
 351static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
 352{
 353        /*
 354         * We do the test below only for project quotas. 'usrquota' and
 355         * 'grpquota' mount options are allowed even without quota feature
 356         * to support legacy quotas in quota files.
 357         */
 358        if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
 359                f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
 360                return -1;
 361        }
 362        if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
 363                        F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
 364                        F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
 365                if (test_opt(sbi, USRQUOTA) &&
 366                                F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
 367                        clear_opt(sbi, USRQUOTA);
 368
 369                if (test_opt(sbi, GRPQUOTA) &&
 370                                F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
 371                        clear_opt(sbi, GRPQUOTA);
 372
 373                if (test_opt(sbi, PRJQUOTA) &&
 374                                F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
 375                        clear_opt(sbi, PRJQUOTA);
 376
 377                if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
 378                                test_opt(sbi, PRJQUOTA)) {
 379                        f2fs_err(sbi, "old and new quota format mixing");
 380                        return -1;
 381                }
 382
 383                if (!F2FS_OPTION(sbi).s_jquota_fmt) {
 384                        f2fs_err(sbi, "journaled quota format not specified");
 385                        return -1;
 386                }
 387        }
 388
 389        if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
 390                f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
 391                F2FS_OPTION(sbi).s_jquota_fmt = 0;
 392        }
 393        return 0;
 394}
 395#endif
 396
 397static int parse_options(struct super_block *sb, char *options)
 398{
 399        struct f2fs_sb_info *sbi = F2FS_SB(sb);
 400        substring_t args[MAX_OPT_ARGS];
 401        unsigned char (*ext)[F2FS_EXTENSION_LEN];
 402        char *p, *name;
 403        int arg = 0, ext_cnt;
 404        kuid_t uid;
 405        kgid_t gid;
 406#ifdef CONFIG_QUOTA
 407        int ret;
 408#endif
 409
 410        if (!options)
 411                return 0;
 412
 413        while ((p = strsep(&options, ",")) != NULL) {
 414                int token;
 415                if (!*p)
 416                        continue;
 417                /*
 418                 * Initialize args struct so we know whether arg was
 419                 * found; some options take optional arguments.
 420                 */
 421                args[0].to = args[0].from = NULL;
 422                token = match_token(p, f2fs_tokens, args);
 423
 424                switch (token) {
 425                case Opt_gc_background:
 426                        name = match_strdup(&args[0]);
 427
 428                        if (!name)
 429                                return -ENOMEM;
 430                        if (strlen(name) == 2 && !strncmp(name, "on", 2)) {
 431                                F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
 432                        } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) {
 433                                F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
 434                        } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) {
 435                                F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
 436                        } else {
 437                                kvfree(name);
 438                                return -EINVAL;
 439                        }
 440                        kvfree(name);
 441                        break;
 442                case Opt_disable_roll_forward:
 443                        set_opt(sbi, DISABLE_ROLL_FORWARD);
 444                        break;
 445                case Opt_norecovery:
 446                        /* this option mounts f2fs with ro */
 447                        set_opt(sbi, NORECOVERY);
 448                        if (!f2fs_readonly(sb))
 449                                return -EINVAL;
 450                        break;
 451                case Opt_discard:
 452                        set_opt(sbi, DISCARD);
 453                        break;
 454                case Opt_nodiscard:
 455                        if (f2fs_sb_has_blkzoned(sbi)) {
 456                                f2fs_warn(sbi, "discard is required for zoned block devices");
 457                                return -EINVAL;
 458                        }
 459                        clear_opt(sbi, DISCARD);
 460                        break;
 461                case Opt_noheap:
 462                        set_opt(sbi, NOHEAP);
 463                        break;
 464                case Opt_heap:
 465                        clear_opt(sbi, NOHEAP);
 466                        break;
 467#ifdef CONFIG_F2FS_FS_XATTR
 468                case Opt_user_xattr:
 469                        set_opt(sbi, XATTR_USER);
 470                        break;
 471                case Opt_nouser_xattr:
 472                        clear_opt(sbi, XATTR_USER);
 473                        break;
 474                case Opt_inline_xattr:
 475                        set_opt(sbi, INLINE_XATTR);
 476                        break;
 477                case Opt_noinline_xattr:
 478                        clear_opt(sbi, INLINE_XATTR);
 479                        break;
 480                case Opt_inline_xattr_size:
 481                        if (args->from && match_int(args, &arg))
 482                                return -EINVAL;
 483                        set_opt(sbi, INLINE_XATTR_SIZE);
 484                        F2FS_OPTION(sbi).inline_xattr_size = arg;
 485                        break;
 486#else
 487                case Opt_user_xattr:
 488                        f2fs_info(sbi, "user_xattr options not supported");
 489                        break;
 490                case Opt_nouser_xattr:
 491                        f2fs_info(sbi, "nouser_xattr options not supported");
 492                        break;
 493                case Opt_inline_xattr:
 494                        f2fs_info(sbi, "inline_xattr options not supported");
 495                        break;
 496                case Opt_noinline_xattr:
 497                        f2fs_info(sbi, "noinline_xattr options not supported");
 498                        break;
 499#endif
 500#ifdef CONFIG_F2FS_FS_POSIX_ACL
 501                case Opt_acl:
 502                        set_opt(sbi, POSIX_ACL);
 503                        break;
 504                case Opt_noacl:
 505                        clear_opt(sbi, POSIX_ACL);
 506                        break;
 507#else
 508                case Opt_acl:
 509                        f2fs_info(sbi, "acl options not supported");
 510                        break;
 511                case Opt_noacl:
 512                        f2fs_info(sbi, "noacl options not supported");
 513                        break;
 514#endif
 515                case Opt_active_logs:
 516                        if (args->from && match_int(args, &arg))
 517                                return -EINVAL;
 518                        if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
 519                                return -EINVAL;
 520                        F2FS_OPTION(sbi).active_logs = arg;
 521                        break;
 522                case Opt_disable_ext_identify:
 523                        set_opt(sbi, DISABLE_EXT_IDENTIFY);
 524                        break;
 525                case Opt_inline_data:
 526                        set_opt(sbi, INLINE_DATA);
 527                        break;
 528                case Opt_inline_dentry:
 529                        set_opt(sbi, INLINE_DENTRY);
 530                        break;
 531                case Opt_noinline_dentry:
 532                        clear_opt(sbi, INLINE_DENTRY);
 533                        break;
 534                case Opt_flush_merge:
 535                        set_opt(sbi, FLUSH_MERGE);
 536                        break;
 537                case Opt_noflush_merge:
 538                        clear_opt(sbi, FLUSH_MERGE);
 539                        break;
 540                case Opt_nobarrier:
 541                        set_opt(sbi, NOBARRIER);
 542                        break;
 543                case Opt_fastboot:
 544                        set_opt(sbi, FASTBOOT);
 545                        break;
 546                case Opt_extent_cache:
 547                        set_opt(sbi, EXTENT_CACHE);
 548                        break;
 549                case Opt_noextent_cache:
 550                        clear_opt(sbi, EXTENT_CACHE);
 551                        break;
 552                case Opt_noinline_data:
 553                        clear_opt(sbi, INLINE_DATA);
 554                        break;
 555                case Opt_data_flush:
 556                        set_opt(sbi, DATA_FLUSH);
 557                        break;
 558                case Opt_reserve_root:
 559                        if (args->from && match_int(args, &arg))
 560                                return -EINVAL;
 561                        if (test_opt(sbi, RESERVE_ROOT)) {
 562                                f2fs_info(sbi, "Preserve previous reserve_root=%u",
 563                                          F2FS_OPTION(sbi).root_reserved_blocks);
 564                        } else {
 565                                F2FS_OPTION(sbi).root_reserved_blocks = arg;
 566                                set_opt(sbi, RESERVE_ROOT);
 567                        }
 568                        break;
 569                case Opt_resuid:
 570                        if (args->from && match_int(args, &arg))
 571                                return -EINVAL;
 572                        uid = make_kuid(current_user_ns(), arg);
 573                        if (!uid_valid(uid)) {
 574                                f2fs_err(sbi, "Invalid uid value %d", arg);
 575                                return -EINVAL;
 576                        }
 577                        F2FS_OPTION(sbi).s_resuid = uid;
 578                        break;
 579                case Opt_resgid:
 580                        if (args->from && match_int(args, &arg))
 581                                return -EINVAL;
 582                        gid = make_kgid(current_user_ns(), arg);
 583                        if (!gid_valid(gid)) {
 584                                f2fs_err(sbi, "Invalid gid value %d", arg);
 585                                return -EINVAL;
 586                        }
 587                        F2FS_OPTION(sbi).s_resgid = gid;
 588                        break;
 589                case Opt_mode:
 590                        name = match_strdup(&args[0]);
 591
 592                        if (!name)
 593                                return -ENOMEM;
 594                        if (strlen(name) == 8 &&
 595                                        !strncmp(name, "adaptive", 8)) {
 596                                if (f2fs_sb_has_blkzoned(sbi)) {
 597                                        f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
 598                                        kvfree(name);
 599                                        return -EINVAL;
 600                                }
 601                                F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
 602                        } else if (strlen(name) == 3 &&
 603                                        !strncmp(name, "lfs", 3)) {
 604                                F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
 605                        } else {
 606                                kvfree(name);
 607                                return -EINVAL;
 608                        }
 609                        kvfree(name);
 610                        break;
 611                case Opt_io_size_bits:
 612                        if (args->from && match_int(args, &arg))
 613                                return -EINVAL;
 614                        if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
 615                                f2fs_warn(sbi, "Not support %d, larger than %d",
 616                                          1 << arg, BIO_MAX_PAGES);
 617                                return -EINVAL;
 618                        }
 619                        F2FS_OPTION(sbi).write_io_size_bits = arg;
 620                        break;
 621#ifdef CONFIG_F2FS_FAULT_INJECTION
 622                case Opt_fault_injection:
 623                        if (args->from && match_int(args, &arg))
 624                                return -EINVAL;
 625                        f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE);
 626                        set_opt(sbi, FAULT_INJECTION);
 627                        break;
 628
 629                case Opt_fault_type:
 630                        if (args->from && match_int(args, &arg))
 631                                return -EINVAL;
 632                        f2fs_build_fault_attr(sbi, 0, arg);
 633                        set_opt(sbi, FAULT_INJECTION);
 634                        break;
 635#else
 636                case Opt_fault_injection:
 637                        f2fs_info(sbi, "fault_injection options not supported");
 638                        break;
 639
 640                case Opt_fault_type:
 641                        f2fs_info(sbi, "fault_type options not supported");
 642                        break;
 643#endif
 644                case Opt_lazytime:
 645                        sb->s_flags |= SB_LAZYTIME;
 646                        break;
 647                case Opt_nolazytime:
 648                        sb->s_flags &= ~SB_LAZYTIME;
 649                        break;
 650#ifdef CONFIG_QUOTA
 651                case Opt_quota:
 652                case Opt_usrquota:
 653                        set_opt(sbi, USRQUOTA);
 654                        break;
 655                case Opt_grpquota:
 656                        set_opt(sbi, GRPQUOTA);
 657                        break;
 658                case Opt_prjquota:
 659                        set_opt(sbi, PRJQUOTA);
 660                        break;
 661                case Opt_usrjquota:
 662                        ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
 663                        if (ret)
 664                                return ret;
 665                        break;
 666                case Opt_grpjquota:
 667                        ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
 668                        if (ret)
 669                                return ret;
 670                        break;
 671                case Opt_prjjquota:
 672                        ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
 673                        if (ret)
 674                                return ret;
 675                        break;
 676                case Opt_offusrjquota:
 677                        ret = f2fs_clear_qf_name(sb, USRQUOTA);
 678                        if (ret)
 679                                return ret;
 680                        break;
 681                case Opt_offgrpjquota:
 682                        ret = f2fs_clear_qf_name(sb, GRPQUOTA);
 683                        if (ret)
 684                                return ret;
 685                        break;
 686                case Opt_offprjjquota:
 687                        ret = f2fs_clear_qf_name(sb, PRJQUOTA);
 688                        if (ret)
 689                                return ret;
 690                        break;
 691                case Opt_jqfmt_vfsold:
 692                        F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
 693                        break;
 694                case Opt_jqfmt_vfsv0:
 695                        F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
 696                        break;
 697                case Opt_jqfmt_vfsv1:
 698                        F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
 699                        break;
 700                case Opt_noquota:
 701                        clear_opt(sbi, QUOTA);
 702                        clear_opt(sbi, USRQUOTA);
 703                        clear_opt(sbi, GRPQUOTA);
 704                        clear_opt(sbi, PRJQUOTA);
 705                        break;
 706#else
 707                case Opt_quota:
 708                case Opt_usrquota:
 709                case Opt_grpquota:
 710                case Opt_prjquota:
 711                case Opt_usrjquota:
 712                case Opt_grpjquota:
 713                case Opt_prjjquota:
 714                case Opt_offusrjquota:
 715                case Opt_offgrpjquota:
 716                case Opt_offprjjquota:
 717                case Opt_jqfmt_vfsold:
 718                case Opt_jqfmt_vfsv0:
 719                case Opt_jqfmt_vfsv1:
 720                case Opt_noquota:
 721                        f2fs_info(sbi, "quota operations not supported");
 722                        break;
 723#endif
 724                case Opt_whint:
 725                        name = match_strdup(&args[0]);
 726                        if (!name)
 727                                return -ENOMEM;
 728                        if (strlen(name) == 10 &&
 729                                        !strncmp(name, "user-based", 10)) {
 730                                F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
 731                        } else if (strlen(name) == 3 &&
 732                                        !strncmp(name, "off", 3)) {
 733                                F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 734                        } else if (strlen(name) == 8 &&
 735                                        !strncmp(name, "fs-based", 8)) {
 736                                F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
 737                        } else {
 738                                kvfree(name);
 739                                return -EINVAL;
 740                        }
 741                        kvfree(name);
 742                        break;
 743                case Opt_alloc:
 744                        name = match_strdup(&args[0]);
 745                        if (!name)
 746                                return -ENOMEM;
 747
 748                        if (strlen(name) == 7 &&
 749                                        !strncmp(name, "default", 7)) {
 750                                F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
 751                        } else if (strlen(name) == 5 &&
 752                                        !strncmp(name, "reuse", 5)) {
 753                                F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
 754                        } else {
 755                                kvfree(name);
 756                                return -EINVAL;
 757                        }
 758                        kvfree(name);
 759                        break;
 760                case Opt_fsync:
 761                        name = match_strdup(&args[0]);
 762                        if (!name)
 763                                return -ENOMEM;
 764                        if (strlen(name) == 5 &&
 765                                        !strncmp(name, "posix", 5)) {
 766                                F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
 767                        } else if (strlen(name) == 6 &&
 768                                        !strncmp(name, "strict", 6)) {
 769                                F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
 770                        } else if (strlen(name) == 9 &&
 771                                        !strncmp(name, "nobarrier", 9)) {
 772                                F2FS_OPTION(sbi).fsync_mode =
 773                                                        FSYNC_MODE_NOBARRIER;
 774                        } else {
 775                                kvfree(name);
 776                                return -EINVAL;
 777                        }
 778                        kvfree(name);
 779                        break;
 780                case Opt_test_dummy_encryption:
 781#ifdef CONFIG_FS_ENCRYPTION
 782                        if (!f2fs_sb_has_encrypt(sbi)) {
 783                                f2fs_err(sbi, "Encrypt feature is off");
 784                                return -EINVAL;
 785                        }
 786
 787                        F2FS_OPTION(sbi).test_dummy_encryption = true;
 788                        f2fs_info(sbi, "Test dummy encryption mode enabled");
 789#else
 790                        f2fs_info(sbi, "Test dummy encryption mount option ignored");
 791#endif
 792                        break;
 793                case Opt_checkpoint_disable_cap_perc:
 794                        if (args->from && match_int(args, &arg))
 795                                return -EINVAL;
 796                        if (arg < 0 || arg > 100)
 797                                return -EINVAL;
 798                        if (arg == 100)
 799                                F2FS_OPTION(sbi).unusable_cap =
 800                                        sbi->user_block_count;
 801                        else
 802                                F2FS_OPTION(sbi).unusable_cap =
 803                                        (sbi->user_block_count / 100) * arg;
 804                        set_opt(sbi, DISABLE_CHECKPOINT);
 805                        break;
 806                case Opt_checkpoint_disable_cap:
 807                        if (args->from && match_int(args, &arg))
 808                                return -EINVAL;
 809                        F2FS_OPTION(sbi).unusable_cap = arg;
 810                        set_opt(sbi, DISABLE_CHECKPOINT);
 811                        break;
 812                case Opt_checkpoint_disable:
 813                        set_opt(sbi, DISABLE_CHECKPOINT);
 814                        break;
 815                case Opt_checkpoint_enable:
 816                        clear_opt(sbi, DISABLE_CHECKPOINT);
 817                        break;
 818                case Opt_compress_algorithm:
 819                        if (!f2fs_sb_has_compression(sbi)) {
 820                                f2fs_err(sbi, "Compression feature if off");
 821                                return -EINVAL;
 822                        }
 823                        name = match_strdup(&args[0]);
 824                        if (!name)
 825                                return -ENOMEM;
 826                        if (strlen(name) == 3 && !strcmp(name, "lzo")) {
 827                                F2FS_OPTION(sbi).compress_algorithm =
 828                                                                COMPRESS_LZO;
 829                        } else if (strlen(name) == 3 &&
 830                                        !strcmp(name, "lz4")) {
 831                                F2FS_OPTION(sbi).compress_algorithm =
 832                                                                COMPRESS_LZ4;
 833                        } else if (strlen(name) == 4 &&
 834                                        !strcmp(name, "zstd")) {
 835                                F2FS_OPTION(sbi).compress_algorithm =
 836                                                                COMPRESS_ZSTD;
 837                        } else {
 838                                kfree(name);
 839                                return -EINVAL;
 840                        }
 841                        kfree(name);
 842                        break;
 843                case Opt_compress_log_size:
 844                        if (!f2fs_sb_has_compression(sbi)) {
 845                                f2fs_err(sbi, "Compression feature is off");
 846                                return -EINVAL;
 847                        }
 848                        if (args->from && match_int(args, &arg))
 849                                return -EINVAL;
 850                        if (arg < MIN_COMPRESS_LOG_SIZE ||
 851                                arg > MAX_COMPRESS_LOG_SIZE) {
 852                                f2fs_err(sbi,
 853                                        "Compress cluster log size is out of range");
 854                                return -EINVAL;
 855                        }
 856                        F2FS_OPTION(sbi).compress_log_size = arg;
 857                        break;
 858                case Opt_compress_extension:
 859                        if (!f2fs_sb_has_compression(sbi)) {
 860                                f2fs_err(sbi, "Compression feature is off");
 861                                return -EINVAL;
 862                        }
 863                        name = match_strdup(&args[0]);
 864                        if (!name)
 865                                return -ENOMEM;
 866
 867                        ext = F2FS_OPTION(sbi).extensions;
 868                        ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
 869
 870                        if (strlen(name) >= F2FS_EXTENSION_LEN ||
 871                                ext_cnt >= COMPRESS_EXT_NUM) {
 872                                f2fs_err(sbi,
 873                                        "invalid extension length/number");
 874                                kfree(name);
 875                                return -EINVAL;
 876                        }
 877
 878                        strcpy(ext[ext_cnt], name);
 879                        F2FS_OPTION(sbi).compress_ext_cnt++;
 880                        kfree(name);
 881                        break;
 882                default:
 883                        f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
 884                                 p);
 885                        return -EINVAL;
 886                }
 887        }
 888#ifdef CONFIG_QUOTA
 889        if (f2fs_check_quota_options(sbi))
 890                return -EINVAL;
 891#else
 892        if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
 893                f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
 894                return -EINVAL;
 895        }
 896        if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
 897                f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
 898                return -EINVAL;
 899        }
 900#endif
 901#ifndef CONFIG_UNICODE
 902        if (f2fs_sb_has_casefold(sbi)) {
 903                f2fs_err(sbi,
 904                        "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
 905                return -EINVAL;
 906        }
 907#endif
 908
 909        if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
 910                f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
 911                         F2FS_IO_SIZE_KB(sbi));
 912                return -EINVAL;
 913        }
 914
 915        if (test_opt(sbi, INLINE_XATTR_SIZE)) {
 916                int min_size, max_size;
 917
 918                if (!f2fs_sb_has_extra_attr(sbi) ||
 919                        !f2fs_sb_has_flexible_inline_xattr(sbi)) {
 920                        f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
 921                        return -EINVAL;
 922                }
 923                if (!test_opt(sbi, INLINE_XATTR)) {
 924                        f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
 925                        return -EINVAL;
 926                }
 927
 928                min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
 929                max_size = MAX_INLINE_XATTR_SIZE;
 930
 931                if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
 932                                F2FS_OPTION(sbi).inline_xattr_size > max_size) {
 933                        f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
 934                                 min_size, max_size);
 935                        return -EINVAL;
 936                }
 937        }
 938
 939        if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
 940                f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n");
 941                return -EINVAL;
 942        }
 943
 944        /* Not pass down write hints if the number of active logs is lesser
 945         * than NR_CURSEG_TYPE.
 946         */
 947        if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE)
 948                F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
 949        return 0;
 950}
 951
 952static struct inode *f2fs_alloc_inode(struct super_block *sb)
 953{
 954        struct f2fs_inode_info *fi;
 955
 956        fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
 957        if (!fi)
 958                return NULL;
 959
 960        init_once((void *) fi);
 961
 962        /* Initialize f2fs-specific inode info */
 963        atomic_set(&fi->dirty_pages, 0);
 964        init_rwsem(&fi->i_sem);
 965        spin_lock_init(&fi->i_size_lock);
 966        INIT_LIST_HEAD(&fi->dirty_list);
 967        INIT_LIST_HEAD(&fi->gdirty_list);
 968        INIT_LIST_HEAD(&fi->inmem_ilist);
 969        INIT_LIST_HEAD(&fi->inmem_pages);
 970        mutex_init(&fi->inmem_lock);
 971        init_rwsem(&fi->i_gc_rwsem[READ]);
 972        init_rwsem(&fi->i_gc_rwsem[WRITE]);
 973        init_rwsem(&fi->i_mmap_sem);
 974        init_rwsem(&fi->i_xattr_sem);
 975
 976        /* Will be used by directory only */
 977        fi->i_dir_level = F2FS_SB(sb)->dir_level;
 978
 979        return &fi->vfs_inode;
 980}
 981
 982static int f2fs_drop_inode(struct inode *inode)
 983{
 984        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 985        int ret;
 986
 987        /*
 988         * during filesystem shutdown, if checkpoint is disabled,
 989         * drop useless meta/node dirty pages.
 990         */
 991        if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
 992                if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 993                        inode->i_ino == F2FS_META_INO(sbi)) {
 994                        trace_f2fs_drop_inode(inode, 1);
 995                        return 1;
 996                }
 997        }
 998
 999        /*
1000         * This is to avoid a deadlock condition like below.
1001         * writeback_single_inode(inode)
1002         *  - f2fs_write_data_page
1003         *    - f2fs_gc -> iput -> evict
1004         *       - inode_wait_for_writeback(inode)
1005         */
1006        if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1007                if (!inode->i_nlink && !is_bad_inode(inode)) {
1008                        /* to avoid evict_inode call simultaneously */
1009                        atomic_inc(&inode->i_count);
1010                        spin_unlock(&inode->i_lock);
1011
1012                        /* some remained atomic pages should discarded */
1013                        if (f2fs_is_atomic_file(inode))
1014                                f2fs_drop_inmem_pages(inode);
1015
1016                        /* should remain fi->extent_tree for writepage */
1017                        f2fs_destroy_extent_node(inode);
1018
1019                        sb_start_intwrite(inode->i_sb);
1020                        f2fs_i_size_write(inode, 0);
1021
1022                        f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1023                                        inode, NULL, 0, DATA);
1024                        truncate_inode_pages_final(inode->i_mapping);
1025
1026                        if (F2FS_HAS_BLOCKS(inode))
1027                                f2fs_truncate(inode);
1028
1029                        sb_end_intwrite(inode->i_sb);
1030
1031                        spin_lock(&inode->i_lock);
1032                        atomic_dec(&inode->i_count);
1033                }
1034                trace_f2fs_drop_inode(inode, 0);
1035                return 0;
1036        }
1037        ret = generic_drop_inode(inode);
1038        if (!ret)
1039                ret = fscrypt_drop_inode(inode);
1040        trace_f2fs_drop_inode(inode, ret);
1041        return ret;
1042}
1043
1044int f2fs_inode_dirtied(struct inode *inode, bool sync)
1045{
1046        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1047        int ret = 0;
1048
1049        spin_lock(&sbi->inode_lock[DIRTY_META]);
1050        if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1051                ret = 1;
1052        } else {
1053                set_inode_flag(inode, FI_DIRTY_INODE);
1054                stat_inc_dirty_inode(sbi, DIRTY_META);
1055        }
1056        if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1057                list_add_tail(&F2FS_I(inode)->gdirty_list,
1058                                &sbi->inode_list[DIRTY_META]);
1059                inc_page_count(sbi, F2FS_DIRTY_IMETA);
1060        }
1061        spin_unlock(&sbi->inode_lock[DIRTY_META]);
1062        return ret;
1063}
1064
1065void f2fs_inode_synced(struct inode *inode)
1066{
1067        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1068
1069        spin_lock(&sbi->inode_lock[DIRTY_META]);
1070        if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1071                spin_unlock(&sbi->inode_lock[DIRTY_META]);
1072                return;
1073        }
1074        if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1075                list_del_init(&F2FS_I(inode)->gdirty_list);
1076                dec_page_count(sbi, F2FS_DIRTY_IMETA);
1077        }
1078        clear_inode_flag(inode, FI_DIRTY_INODE);
1079        clear_inode_flag(inode, FI_AUTO_RECOVER);
1080        stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1081        spin_unlock(&sbi->inode_lock[DIRTY_META]);
1082}
1083
1084/*
1085 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1086 *
1087 * We should call set_dirty_inode to write the dirty inode through write_inode.
1088 */
1089static void f2fs_dirty_inode(struct inode *inode, int flags)
1090{
1091        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1092
1093        if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1094                        inode->i_ino == F2FS_META_INO(sbi))
1095                return;
1096
1097        if (flags == I_DIRTY_TIME)
1098                return;
1099
1100        if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1101                clear_inode_flag(inode, FI_AUTO_RECOVER);
1102
1103        f2fs_inode_dirtied(inode, false);
1104}
1105
1106static void f2fs_free_inode(struct inode *inode)
1107{
1108        fscrypt_free_inode(inode);
1109        kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1110}
1111
1112static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1113{
1114        percpu_counter_destroy(&sbi->alloc_valid_block_count);
1115        percpu_counter_destroy(&sbi->total_valid_inode_count);
1116}
1117
1118static void destroy_device_list(struct f2fs_sb_info *sbi)
1119{
1120        int i;
1121
1122        for (i = 0; i < sbi->s_ndevs; i++) {
1123                blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1124#ifdef CONFIG_BLK_DEV_ZONED
1125                kvfree(FDEV(i).blkz_seq);
1126#endif
1127        }
1128        kvfree(sbi->devs);
1129}
1130
1131static void f2fs_put_super(struct super_block *sb)
1132{
1133        struct f2fs_sb_info *sbi = F2FS_SB(sb);
1134        int i;
1135        bool dropped;
1136
1137        f2fs_quota_off_umount(sb);
1138
1139        /* prevent remaining shrinker jobs */
1140        mutex_lock(&sbi->umount_mutex);
1141
1142        /*
1143         * We don't need to do checkpoint when superblock is clean.
1144         * But, the previous checkpoint was not done by umount, it needs to do
1145         * clean checkpoint again.
1146         */
1147        if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1148                        !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1149                struct cp_control cpc = {
1150                        .reason = CP_UMOUNT,
1151                };
1152                f2fs_write_checkpoint(sbi, &cpc);
1153        }
1154
1155        /* be sure to wait for any on-going discard commands */
1156        dropped = f2fs_issue_discard_timeout(sbi);
1157
1158        if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1159                                        !sbi->discard_blks && !dropped) {
1160                struct cp_control cpc = {
1161                        .reason = CP_UMOUNT | CP_TRIMMED,
1162                };
1163                f2fs_write_checkpoint(sbi, &cpc);
1164        }
1165
1166        /*
1167         * normally superblock is clean, so we need to release this.
1168         * In addition, EIO will skip do checkpoint, we need this as well.
1169         */
1170        f2fs_release_ino_entry(sbi, true);
1171
1172        f2fs_leave_shrinker(sbi);
1173        mutex_unlock(&sbi->umount_mutex);
1174
1175        /* our cp_error case, we can wait for any writeback page */
1176        f2fs_flush_merged_writes(sbi);
1177
1178        f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1179
1180        f2fs_bug_on(sbi, sbi->fsync_node_num);
1181
1182        iput(sbi->node_inode);
1183        sbi->node_inode = NULL;
1184
1185        iput(sbi->meta_inode);
1186        sbi->meta_inode = NULL;
1187
1188        /*
1189         * iput() can update stat information, if f2fs_write_checkpoint()
1190         * above failed with error.
1191         */
1192        f2fs_destroy_stats(sbi);
1193
1194        /* destroy f2fs internal modules */
1195        f2fs_destroy_node_manager(sbi);
1196        f2fs_destroy_segment_manager(sbi);
1197
1198        f2fs_destroy_post_read_wq(sbi);
1199
1200        kvfree(sbi->ckpt);
1201
1202        f2fs_unregister_sysfs(sbi);
1203
1204        sb->s_fs_info = NULL;
1205        if (sbi->s_chksum_driver)
1206                crypto_free_shash(sbi->s_chksum_driver);
1207        kvfree(sbi->raw_super);
1208
1209        destroy_device_list(sbi);
1210        f2fs_destroy_xattr_caches(sbi);
1211        mempool_destroy(sbi->write_io_dummy);
1212#ifdef CONFIG_QUOTA
1213        for (i = 0; i < MAXQUOTAS; i++)
1214                kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1215#endif
1216        destroy_percpu_info(sbi);
1217        for (i = 0; i < NR_PAGE_TYPE; i++)
1218                kvfree(sbi->write_io[i]);
1219#ifdef CONFIG_UNICODE
1220        utf8_unload(sbi->s_encoding);
1221#endif
1222        kvfree(sbi);
1223}
1224
1225int f2fs_sync_fs(struct super_block *sb, int sync)
1226{
1227        struct f2fs_sb_info *sbi = F2FS_SB(sb);
1228        int err = 0;
1229
1230        if (unlikely(f2fs_cp_error(sbi)))
1231                return 0;
1232        if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1233                return 0;
1234
1235        trace_f2fs_sync_fs(sb, sync);
1236
1237        if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1238                return -EAGAIN;
1239
1240        if (sync) {
1241                struct cp_control cpc;
1242
1243                cpc.reason = __get_cp_reason(sbi);
1244
1245                down_write(&sbi->gc_lock);
1246                err = f2fs_write_checkpoint(sbi, &cpc);
1247                up_write(&sbi->gc_lock);
1248        }
1249        f2fs_trace_ios(NULL, 1);
1250
1251        return err;
1252}
1253
1254static int f2fs_freeze(struct super_block *sb)
1255{
1256        if (f2fs_readonly(sb))
1257                return 0;
1258
1259        /* IO error happened before */
1260        if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1261                return -EIO;
1262
1263        /* must be clean, since sync_filesystem() was already called */
1264        if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1265                return -EINVAL;
1266        return 0;
1267}
1268
1269static int f2fs_unfreeze(struct super_block *sb)
1270{
1271        return 0;
1272}
1273
1274#ifdef CONFIG_QUOTA
1275static int f2fs_statfs_project(struct super_block *sb,
1276                                kprojid_t projid, struct kstatfs *buf)
1277{
1278        struct kqid qid;
1279        struct dquot *dquot;
1280        u64 limit;
1281        u64 curblock;
1282
1283        qid = make_kqid_projid(projid);
1284        dquot = dqget(sb, qid);
1285        if (IS_ERR(dquot))
1286                return PTR_ERR(dquot);
1287        spin_lock(&dquot->dq_dqb_lock);
1288
1289        limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1290                                        dquot->dq_dqb.dqb_bhardlimit);
1291        if (limit)
1292                limit >>= sb->s_blocksize_bits;
1293
1294        if (limit && buf->f_blocks > limit) {
1295                curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
1296                buf->f_blocks = limit;
1297                buf->f_bfree = buf->f_bavail =
1298                        (buf->f_blocks > curblock) ?
1299                         (buf->f_blocks - curblock) : 0;
1300        }
1301
1302        limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1303                                        dquot->dq_dqb.dqb_ihardlimit);
1304
1305        if (limit && buf->f_files > limit) {
1306                buf->f_files = limit;
1307                buf->f_ffree =
1308                        (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1309                         (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1310        }
1311
1312        spin_unlock(&dquot->dq_dqb_lock);
1313        dqput(dquot);
1314        return 0;
1315}
1316#endif
1317
1318static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1319{
1320        struct super_block *sb = dentry->d_sb;
1321        struct f2fs_sb_info *sbi = F2FS_SB(sb);
1322        u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1323        block_t total_count, user_block_count, start_count;
1324        u64 avail_node_count;
1325
1326        total_count = le64_to_cpu(sbi->raw_super->block_count);
1327        user_block_count = sbi->user_block_count;
1328        start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1329        buf->f_type = F2FS_SUPER_MAGIC;
1330        buf->f_bsize = sbi->blocksize;
1331
1332        buf->f_blocks = total_count - start_count;
1333        buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1334                                                sbi->current_reserved_blocks;
1335
1336        spin_lock(&sbi->stat_lock);
1337        if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1338                buf->f_bfree = 0;
1339        else
1340                buf->f_bfree -= sbi->unusable_block_count;
1341        spin_unlock(&sbi->stat_lock);
1342
1343        if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1344                buf->f_bavail = buf->f_bfree -
1345                                F2FS_OPTION(sbi).root_reserved_blocks;
1346        else
1347                buf->f_bavail = 0;
1348
1349        avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1350
1351        if (avail_node_count > user_block_count) {
1352                buf->f_files = user_block_count;
1353                buf->f_ffree = buf->f_bavail;
1354        } else {
1355                buf->f_files = avail_node_count;
1356                buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1357                                        buf->f_bavail);
1358        }
1359
1360        buf->f_namelen = F2FS_NAME_LEN;
1361        buf->f_fsid.val[0] = (u32)id;
1362        buf->f_fsid.val[1] = (u32)(id >> 32);
1363
1364#ifdef CONFIG_QUOTA
1365        if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1366                        sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1367                f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1368        }
1369#endif
1370        return 0;
1371}
1372
1373static inline void f2fs_show_quota_options(struct seq_file *seq,
1374                                           struct super_block *sb)
1375{
1376#ifdef CONFIG_QUOTA
1377        struct f2fs_sb_info *sbi = F2FS_SB(sb);
1378
1379        if (F2FS_OPTION(sbi).s_jquota_fmt) {
1380                char *fmtname = "";
1381
1382                switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1383                case QFMT_VFS_OLD:
1384                        fmtname = "vfsold";
1385                        break;
1386                case QFMT_VFS_V0:
1387                        fmtname = "vfsv0";
1388                        break;
1389                case QFMT_VFS_V1:
1390                        fmtname = "vfsv1";
1391                        break;
1392                }
1393                seq_printf(seq, ",jqfmt=%s", fmtname);
1394        }
1395
1396        if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1397                seq_show_option(seq, "usrjquota",
1398                        F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1399
1400        if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1401                seq_show_option(seq, "grpjquota",
1402                        F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1403
1404        if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1405                seq_show_option(seq, "prjjquota",
1406                        F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1407#endif
1408}
1409
1410static inline void f2fs_show_compress_options(struct seq_file *seq,
1411                                                        struct super_block *sb)
1412{
1413        struct f2fs_sb_info *sbi = F2FS_SB(sb);
1414        char *algtype = "";
1415        int i;
1416
1417        if (!f2fs_sb_has_compression(sbi))
1418                return;
1419
1420        switch (F2FS_OPTION(sbi).compress_algorithm) {
1421        case COMPRESS_LZO:
1422                algtype = "lzo";
1423                break;
1424        case COMPRESS_LZ4:
1425                algtype = "lz4";
1426                break;
1427        case COMPRESS_ZSTD:
1428                algtype = "zstd";
1429                break;
1430        }
1431        seq_printf(seq, ",compress_algorithm=%s", algtype);
1432
1433        seq_printf(seq, ",compress_log_size=%u",
1434                        F2FS_OPTION(sbi).compress_log_size);
1435
1436        for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1437                seq_printf(seq, ",compress_extension=%s",
1438                        F2FS_OPTION(sbi).extensions[i]);
1439        }
1440}
1441
1442static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1443{
1444        struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1445
1446        if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1447                seq_printf(seq, ",background_gc=%s", "sync");
1448        else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1449                seq_printf(seq, ",background_gc=%s", "on");
1450        else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1451                seq_printf(seq, ",background_gc=%s", "off");
1452
1453        if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1454                seq_puts(seq, ",disable_roll_forward");
1455        if (test_opt(sbi, NORECOVERY))
1456                seq_puts(seq, ",norecovery");
1457        if (test_opt(sbi, DISCARD))
1458                seq_puts(seq, ",discard");
1459        else
1460                seq_puts(seq, ",nodiscard");
1461        if (test_opt(sbi, NOHEAP))
1462                seq_puts(seq, ",no_heap");
1463        else
1464                seq_puts(seq, ",heap");
1465#ifdef CONFIG_F2FS_FS_XATTR
1466        if (test_opt(sbi, XATTR_USER))
1467                seq_puts(seq, ",user_xattr");
1468        else
1469                seq_puts(seq, ",nouser_xattr");
1470        if (test_opt(sbi, INLINE_XATTR))
1471                seq_puts(seq, ",inline_xattr");
1472        else
1473                seq_puts(seq, ",noinline_xattr");
1474        if (test_opt(sbi, INLINE_XATTR_SIZE))
1475                seq_printf(seq, ",inline_xattr_size=%u",
1476                                        F2FS_OPTION(sbi).inline_xattr_size);
1477#endif
1478#ifdef CONFIG_F2FS_FS_POSIX_ACL
1479        if (test_opt(sbi, POSIX_ACL))
1480                seq_puts(seq, ",acl");
1481        else
1482                seq_puts(seq, ",noacl");
1483#endif
1484        if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1485                seq_puts(seq, ",disable_ext_identify");
1486        if (test_opt(sbi, INLINE_DATA))
1487                seq_puts(seq, ",inline_data");
1488        else
1489                seq_puts(seq, ",noinline_data");
1490        if (test_opt(sbi, INLINE_DENTRY))
1491                seq_puts(seq, ",inline_dentry");
1492        else
1493                seq_puts(seq, ",noinline_dentry");
1494        if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1495                seq_puts(seq, ",flush_merge");
1496        if (test_opt(sbi, NOBARRIER))
1497                seq_puts(seq, ",nobarrier");
1498        if (test_opt(sbi, FASTBOOT))
1499                seq_puts(seq, ",fastboot");
1500        if (test_opt(sbi, EXTENT_CACHE))
1501                seq_puts(seq, ",extent_cache");
1502        else
1503                seq_puts(seq, ",noextent_cache");
1504        if (test_opt(sbi, DATA_FLUSH))
1505                seq_puts(seq, ",data_flush");
1506
1507        seq_puts(seq, ",mode=");
1508        if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1509                seq_puts(seq, "adaptive");
1510        else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1511                seq_puts(seq, "lfs");
1512        seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1513        if (test_opt(sbi, RESERVE_ROOT))
1514                seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1515                                F2FS_OPTION(sbi).root_reserved_blocks,
1516                                from_kuid_munged(&init_user_ns,
1517                                        F2FS_OPTION(sbi).s_resuid),
1518                                from_kgid_munged(&init_user_ns,
1519                                        F2FS_OPTION(sbi).s_resgid));
1520        if (F2FS_IO_SIZE_BITS(sbi))
1521                seq_printf(seq, ",io_bits=%u",
1522                                F2FS_OPTION(sbi).write_io_size_bits);
1523#ifdef CONFIG_F2FS_FAULT_INJECTION
1524        if (test_opt(sbi, FAULT_INJECTION)) {
1525                seq_printf(seq, ",fault_injection=%u",
1526                                F2FS_OPTION(sbi).fault_info.inject_rate);
1527                seq_printf(seq, ",fault_type=%u",
1528                                F2FS_OPTION(sbi).fault_info.inject_type);
1529        }
1530#endif
1531#ifdef CONFIG_QUOTA
1532        if (test_opt(sbi, QUOTA))
1533                seq_puts(seq, ",quota");
1534        if (test_opt(sbi, USRQUOTA))
1535                seq_puts(seq, ",usrquota");
1536        if (test_opt(sbi, GRPQUOTA))
1537                seq_puts(seq, ",grpquota");
1538        if (test_opt(sbi, PRJQUOTA))
1539                seq_puts(seq, ",prjquota");
1540#endif
1541        f2fs_show_quota_options(seq, sbi->sb);
1542        if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1543                seq_printf(seq, ",whint_mode=%s", "user-based");
1544        else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1545                seq_printf(seq, ",whint_mode=%s", "fs-based");
1546#ifdef CONFIG_FS_ENCRYPTION
1547        if (F2FS_OPTION(sbi).test_dummy_encryption)
1548                seq_puts(seq, ",test_dummy_encryption");
1549#endif
1550
1551        if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1552                seq_printf(seq, ",alloc_mode=%s", "default");
1553        else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1554                seq_printf(seq, ",alloc_mode=%s", "reuse");
1555
1556        if (test_opt(sbi, DISABLE_CHECKPOINT))
1557                seq_printf(seq, ",checkpoint=disable:%u",
1558                                F2FS_OPTION(sbi).unusable_cap);
1559        if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1560                seq_printf(seq, ",fsync_mode=%s", "posix");
1561        else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1562                seq_printf(seq, ",fsync_mode=%s", "strict");
1563        else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1564                seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1565
1566        f2fs_show_compress_options(seq, sbi->sb);
1567        return 0;
1568}
1569
1570static void default_options(struct f2fs_sb_info *sbi)
1571{
1572        /* init some FS parameters */
1573        F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE;
1574        F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1575        F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1576        F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1577        F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1578        F2FS_OPTION(sbi).test_dummy_encryption = false;
1579        F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1580        F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1581        F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
1582        F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
1583        F2FS_OPTION(sbi).compress_ext_cnt = 0;
1584        F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
1585
1586        set_opt(sbi, INLINE_XATTR);
1587        set_opt(sbi, INLINE_DATA);
1588        set_opt(sbi, INLINE_DENTRY);
1589        set_opt(sbi, EXTENT_CACHE);
1590        set_opt(sbi, NOHEAP);
1591        clear_opt(sbi, DISABLE_CHECKPOINT);
1592        F2FS_OPTION(sbi).unusable_cap = 0;
1593        sbi->sb->s_flags |= SB_LAZYTIME;
1594        set_opt(sbi, FLUSH_MERGE);
1595        set_opt(sbi, DISCARD);
1596        if (f2fs_sb_has_blkzoned(sbi))
1597                F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
1598        else
1599                F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
1600
1601#ifdef CONFIG_F2FS_FS_XATTR
1602        set_opt(sbi, XATTR_USER);
1603#endif
1604#ifdef CONFIG_F2FS_FS_POSIX_ACL
1605        set_opt(sbi, POSIX_ACL);
1606#endif
1607
1608        f2fs_build_fault_attr(sbi, 0, 0);
1609}
1610
1611#ifdef CONFIG_QUOTA
1612static int f2fs_enable_quotas(struct super_block *sb);
1613#endif
1614
1615static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1616{
1617        unsigned int s_flags = sbi->sb->s_flags;
1618        struct cp_control cpc;
1619        int err = 0;
1620        int ret;
1621        block_t unusable;
1622
1623        if (s_flags & SB_RDONLY) {
1624                f2fs_err(sbi, "checkpoint=disable on readonly fs");
1625                return -EINVAL;
1626        }
1627        sbi->sb->s_flags |= SB_ACTIVE;
1628
1629        f2fs_update_time(sbi, DISABLE_TIME);
1630
1631        while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1632                down_write(&sbi->gc_lock);
1633                err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1634                if (err == -ENODATA) {
1635                        err = 0;
1636                        break;
1637                }
1638                if (err && err != -EAGAIN)
1639                        break;
1640        }
1641
1642        ret = sync_filesystem(sbi->sb);
1643        if (ret || err) {
1644                err = ret ? ret: err;
1645                goto restore_flag;
1646        }
1647
1648        unusable = f2fs_get_unusable_blocks(sbi);
1649        if (f2fs_disable_cp_again(sbi, unusable)) {
1650                err = -EAGAIN;
1651                goto restore_flag;
1652        }
1653
1654        down_write(&sbi->gc_lock);
1655        cpc.reason = CP_PAUSE;
1656        set_sbi_flag(sbi, SBI_CP_DISABLED);
1657        err = f2fs_write_checkpoint(sbi, &cpc);
1658        if (err)
1659                goto out_unlock;
1660
1661        spin_lock(&sbi->stat_lock);
1662        sbi->unusable_block_count = unusable;
1663        spin_unlock(&sbi->stat_lock);
1664
1665out_unlock:
1666        up_write(&sbi->gc_lock);
1667restore_flag:
1668        sbi->sb->s_flags = s_flags;     /* Restore SB_RDONLY status */
1669        return err;
1670}
1671
1672static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1673{
1674        down_write(&sbi->gc_lock);
1675        f2fs_dirty_to_prefree(sbi);
1676
1677        clear_sbi_flag(sbi, SBI_CP_DISABLED);
1678        set_sbi_flag(sbi, SBI_IS_DIRTY);
1679        up_write(&sbi->gc_lock);
1680
1681        f2fs_sync_fs(sbi->sb, 1);
1682}
1683
1684static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1685{
1686        struct f2fs_sb_info *sbi = F2FS_SB(sb);
1687        struct f2fs_mount_info org_mount_opt;
1688        unsigned long old_sb_flags;
1689        int err;
1690        bool need_restart_gc = false;
1691        bool need_stop_gc = false;
1692        bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1693        bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1694        bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1695        bool checkpoint_changed;
1696#ifdef CONFIG_QUOTA
1697        int i, j;
1698#endif
1699
1700        /*
1701         * Save the old mount options in case we
1702         * need to restore them.
1703         */
1704        org_mount_opt = sbi->mount_opt;
1705        old_sb_flags = sb->s_flags;
1706
1707#ifdef CONFIG_QUOTA
1708        org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1709        for (i = 0; i < MAXQUOTAS; i++) {
1710                if (F2FS_OPTION(sbi).s_qf_names[i]) {
1711                        org_mount_opt.s_qf_names[i] =
1712                                kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1713                                GFP_KERNEL);
1714                        if (!org_mount_opt.s_qf_names[i]) {
1715                                for (j = 0; j < i; j++)
1716                                        kvfree(org_mount_opt.s_qf_names[j]);
1717                                return -ENOMEM;
1718                        }
1719                } else {
1720                        org_mount_opt.s_qf_names[i] = NULL;
1721                }
1722        }
1723#endif
1724
1725        /* recover superblocks we couldn't write due to previous RO mount */
1726        if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1727                err = f2fs_commit_super(sbi, false);
1728                f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1729                          err);
1730                if (!err)
1731                        clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1732        }
1733
1734        default_options(sbi);
1735
1736        /* parse mount options */
1737        err = parse_options(sb, data);
1738        if (err)
1739                goto restore_opts;
1740        checkpoint_changed =
1741                        disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1742
1743        /*
1744         * Previous and new state of filesystem is RO,
1745         * so skip checking GC and FLUSH_MERGE conditions.
1746         */
1747        if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1748                goto skip;
1749
1750#ifdef CONFIG_QUOTA
1751        if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1752                err = dquot_suspend(sb, -1);
1753                if (err < 0)
1754                        goto restore_opts;
1755        } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1756                /* dquot_resume needs RW */
1757                sb->s_flags &= ~SB_RDONLY;
1758                if (sb_any_quota_suspended(sb)) {
1759                        dquot_resume(sb, -1);
1760                } else if (f2fs_sb_has_quota_ino(sbi)) {
1761                        err = f2fs_enable_quotas(sb);
1762                        if (err)
1763                                goto restore_opts;
1764                }
1765        }
1766#endif
1767        /* disallow enable/disable extent_cache dynamically */
1768        if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1769                err = -EINVAL;
1770                f2fs_warn(sbi, "switch extent_cache option is not allowed");
1771                goto restore_opts;
1772        }
1773
1774        if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1775                err = -EINVAL;
1776                f2fs_warn(sbi, "switch io_bits option is not allowed");
1777                goto restore_opts;
1778        }
1779
1780        if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1781                err = -EINVAL;
1782                f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1783                goto restore_opts;
1784        }
1785
1786        /*
1787         * We stop the GC thread if FS is mounted as RO
1788         * or if background_gc = off is passed in mount
1789         * option. Also sync the filesystem.
1790         */
1791        if ((*flags & SB_RDONLY) ||
1792                        F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) {
1793                if (sbi->gc_thread) {
1794                        f2fs_stop_gc_thread(sbi);
1795                        need_restart_gc = true;
1796                }
1797        } else if (!sbi->gc_thread) {
1798                err = f2fs_start_gc_thread(sbi);
1799                if (err)
1800                        goto restore_opts;
1801                need_stop_gc = true;
1802        }
1803
1804        if (*flags & SB_RDONLY ||
1805                F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1806                writeback_inodes_sb(sb, WB_REASON_SYNC);
1807                sync_inodes_sb(sb);
1808
1809                set_sbi_flag(sbi, SBI_IS_DIRTY);
1810                set_sbi_flag(sbi, SBI_IS_CLOSE);
1811                f2fs_sync_fs(sb, 1);
1812                clear_sbi_flag(sbi, SBI_IS_CLOSE);
1813        }
1814
1815        if (checkpoint_changed) {
1816                if (test_opt(sbi, DISABLE_CHECKPOINT)) {
1817                        err = f2fs_disable_checkpoint(sbi);
1818                        if (err)
1819                                goto restore_gc;
1820                } else {
1821                        f2fs_enable_checkpoint(sbi);
1822                }
1823        }
1824
1825        /*
1826         * We stop issue flush thread if FS is mounted as RO
1827         * or if flush_merge is not passed in mount option.
1828         */
1829        if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
1830                clear_opt(sbi, FLUSH_MERGE);
1831                f2fs_destroy_flush_cmd_control(sbi, false);
1832        } else {
1833                err = f2fs_create_flush_cmd_control(sbi);
1834                if (err)
1835                        goto restore_gc;
1836        }
1837skip:
1838#ifdef CONFIG_QUOTA
1839        /* Release old quota file names */
1840        for (i = 0; i < MAXQUOTAS; i++)
1841                kvfree(org_mount_opt.s_qf_names[i]);
1842#endif
1843        /* Update the POSIXACL Flag */
1844        sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1845                (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
1846
1847        limit_reserve_root(sbi);
1848        *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
1849        return 0;
1850restore_gc:
1851        if (need_restart_gc) {
1852                if (f2fs_start_gc_thread(sbi))
1853                        f2fs_warn(sbi, "background gc thread has stopped");
1854        } else if (need_stop_gc) {
1855                f2fs_stop_gc_thread(sbi);
1856        }
1857restore_opts:
1858#ifdef CONFIG_QUOTA
1859        F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
1860        for (i = 0; i < MAXQUOTAS; i++) {
1861                kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
1862                F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
1863        }
1864#endif
1865        sbi->mount_opt = org_mount_opt;
1866        sb->s_flags = old_sb_flags;
1867        return err;
1868}
1869
1870#ifdef CONFIG_QUOTA
1871/* Read data from quotafile */
1872static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
1873                               size_t len, loff_t off)
1874{
1875        struct inode *inode = sb_dqopt(sb)->files[type];
1876        struct address_space *mapping = inode->i_mapping;
1877        block_t blkidx = F2FS_BYTES_TO_BLK(off);
1878        int offset = off & (sb->s_blocksize - 1);
1879        int tocopy;
1880        size_t toread;
1881        loff_t i_size = i_size_read(inode);
1882        struct page *page;
1883        char *kaddr;
1884
1885        if (off > i_size)
1886                return 0;
1887
1888        if (off + len > i_size)
1889                len = i_size - off;
1890        toread = len;
1891        while (toread > 0) {
1892                tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1893repeat:
1894                page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
1895                if (IS_ERR(page)) {
1896                        if (PTR_ERR(page) == -ENOMEM) {
1897                                congestion_wait(BLK_RW_ASYNC,
1898                                                DEFAULT_IO_TIMEOUT);
1899                                goto repeat;
1900                        }
1901                        set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1902                        return PTR_ERR(page);
1903                }
1904
1905                lock_page(page);
1906
1907                if (unlikely(page->mapping != mapping)) {
1908                        f2fs_put_page(page, 1);
1909                        goto repeat;
1910                }
1911                if (unlikely(!PageUptodate(page))) {
1912                        f2fs_put_page(page, 1);
1913                        set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1914                        return -EIO;
1915                }
1916
1917                kaddr = kmap_atomic(page);
1918                memcpy(data, kaddr + offset, tocopy);
1919                kunmap_atomic(kaddr);
1920                f2fs_put_page(page, 1);
1921
1922                offset = 0;
1923                toread -= tocopy;
1924                data += tocopy;
1925                blkidx++;
1926        }
1927        return len;
1928}
1929
1930/* Write to quotafile */
1931static ssize_t f2fs_quota_write(struct super_block *sb, int type,
1932                                const char *data, size_t len, loff_t off)
1933{
1934        struct inode *inode = sb_dqopt(sb)->files[type];
1935        struct address_space *mapping = inode->i_mapping;
1936        const struct address_space_operations *a_ops = mapping->a_ops;
1937        int offset = off & (sb->s_blocksize - 1);
1938        size_t towrite = len;
1939        struct page *page;
1940        void *fsdata = NULL;
1941        char *kaddr;
1942        int err = 0;
1943        int tocopy;
1944
1945        while (towrite > 0) {
1946                tocopy = min_t(unsigned long, sb->s_blocksize - offset,
1947                                                                towrite);
1948retry:
1949                err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
1950                                                        &page, &fsdata);
1951                if (unlikely(err)) {
1952                        if (err == -ENOMEM) {
1953                                congestion_wait(BLK_RW_ASYNC,
1954                                                DEFAULT_IO_TIMEOUT);
1955                                goto retry;
1956                        }
1957                        set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
1958                        break;
1959                }
1960
1961                kaddr = kmap_atomic(page);
1962                memcpy(kaddr + offset, data, tocopy);
1963                kunmap_atomic(kaddr);
1964                flush_dcache_page(page);
1965
1966                a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
1967                                                page, fsdata);
1968                offset = 0;
1969                towrite -= tocopy;
1970                off += tocopy;
1971                data += tocopy;
1972                cond_resched();
1973        }
1974
1975        if (len == towrite)
1976                return err;
1977        inode->i_mtime = inode->i_ctime = current_time(inode);
1978        f2fs_mark_inode_dirty_sync(inode, false);
1979        return len - towrite;
1980}
1981
1982static struct dquot **f2fs_get_dquots(struct inode *inode)
1983{
1984        return F2FS_I(inode)->i_dquot;
1985}
1986
1987static qsize_t *f2fs_get_reserved_space(struct inode *inode)
1988{
1989        return &F2FS_I(inode)->i_reserved_quota;
1990}
1991
1992static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
1993{
1994        if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
1995                f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
1996                return 0;
1997        }
1998
1999        return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2000                                        F2FS_OPTION(sbi).s_jquota_fmt, type);
2001}
2002
2003int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2004{
2005        int enabled = 0;
2006        int i, err;
2007
2008        if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2009                err = f2fs_enable_quotas(sbi->sb);
2010                if (err) {
2011                        f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2012                        return 0;
2013                }
2014                return 1;
2015        }
2016
2017        for (i = 0; i < MAXQUOTAS; i++) {
2018                if (F2FS_OPTION(sbi).s_qf_names[i]) {
2019                        err = f2fs_quota_on_mount(sbi, i);
2020                        if (!err) {
2021                                enabled = 1;
2022                                continue;
2023                        }
2024                        f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2025                                 err, i);
2026                }
2027        }
2028        return enabled;
2029}
2030
2031static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2032                             unsigned int flags)
2033{
2034        struct inode *qf_inode;
2035        unsigned long qf_inum;
2036        int err;
2037
2038        BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2039
2040        qf_inum = f2fs_qf_ino(sb, type);
2041        if (!qf_inum)
2042                return -EPERM;
2043
2044        qf_inode = f2fs_iget(sb, qf_inum);
2045        if (IS_ERR(qf_inode)) {
2046                f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2047                return PTR_ERR(qf_inode);
2048        }
2049
2050        /* Don't account quota for quota files to avoid recursion */
2051        qf_inode->i_flags |= S_NOQUOTA;
2052        err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2053        iput(qf_inode);
2054        return err;
2055}
2056
2057static int f2fs_enable_quotas(struct super_block *sb)
2058{
2059        struct f2fs_sb_info *sbi = F2FS_SB(sb);
2060        int type, err = 0;
2061        unsigned long qf_inum;
2062        bool quota_mopt[MAXQUOTAS] = {
2063                test_opt(sbi, USRQUOTA),
2064                test_opt(sbi, GRPQUOTA),
2065                test_opt(sbi, PRJQUOTA),
2066        };
2067
2068        if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2069                f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2070                return 0;
2071        }
2072
2073        sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2074
2075        for (type = 0; type < MAXQUOTAS; type++) {
2076                qf_inum = f2fs_qf_ino(sb, type);
2077                if (qf_inum) {
2078                        err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2079                                DQUOT_USAGE_ENABLED |
2080                                (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2081                        if (err) {
2082                                f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2083                                         type, err);
2084                                for (type--; type >= 0; type--)
2085                                        dquot_quota_off(sb, type);
2086                                set_sbi_flag(F2FS_SB(sb),
2087                                                SBI_QUOTA_NEED_REPAIR);
2088                                return err;
2089                        }
2090                }
2091        }
2092        return 0;
2093}
2094
2095int f2fs_quota_sync(struct super_block *sb, int type)
2096{
2097        struct f2fs_sb_info *sbi = F2FS_SB(sb);
2098        struct quota_info *dqopt = sb_dqopt(sb);
2099        int cnt;
2100        int ret;
2101
2102        /*
2103         * do_quotactl
2104         *  f2fs_quota_sync
2105         *  down_read(quota_sem)
2106         *  dquot_writeback_dquots()
2107         *  f2fs_dquot_commit
2108         *                            block_operation
2109         *                            down_read(quota_sem)
2110         */
2111        f2fs_lock_op(sbi);
2112
2113        down_read(&sbi->quota_sem);
2114        ret = dquot_writeback_dquots(sb, type);
2115        if (ret)
2116                goto out;
2117
2118        /*
2119         * Now when everything is written we can discard the pagecache so
2120         * that userspace sees the changes.
2121         */
2122        for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2123                struct address_space *mapping;
2124
2125                if (type != -1 && cnt != type)
2126                        continue;
2127                if (!sb_has_quota_active(sb, cnt))
2128                        continue;
2129
2130                mapping = dqopt->files[cnt]->i_mapping;
2131
2132                ret = filemap_fdatawrite(mapping);
2133                if (ret)
2134                        goto out;
2135
2136                /* if we are using journalled quota */
2137                if (is_journalled_quota(sbi))
2138                        continue;
2139
2140                ret = filemap_fdatawait(mapping);
2141                if (ret)
2142                        set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2143
2144                inode_lock(dqopt->files[cnt]);
2145                truncate_inode_pages(&dqopt->files[cnt]->i_data, 0);
2146                inode_unlock(dqopt->files[cnt]);
2147        }
2148out:
2149        if (ret)
2150                set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2151        up_read(&sbi->quota_sem);
2152        f2fs_unlock_op(sbi);
2153        return ret;
2154}
2155
2156static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2157                                                        const struct path *path)
2158{
2159        struct inode *inode;
2160        int err;
2161
2162        /* if quota sysfile exists, deny enabling quota with specific file */
2163        if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2164                f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2165                return -EBUSY;
2166        }
2167
2168        err = f2fs_quota_sync(sb, type);
2169        if (err)
2170                return err;
2171
2172        err = dquot_quota_on(sb, type, format_id, path);
2173        if (err)
2174                return err;
2175
2176        inode = d_inode(path->dentry);
2177
2178        inode_lock(inode);
2179        F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2180        f2fs_set_inode_flags(inode);
2181        inode_unlock(inode);
2182        f2fs_mark_inode_dirty_sync(inode, false);
2183
2184        return 0;
2185}
2186
2187static int __f2fs_quota_off(struct super_block *sb, int type)
2188{
2189        struct inode *inode = sb_dqopt(sb)->files[type];
2190        int err;
2191
2192        if (!inode || !igrab(inode))
2193                return dquot_quota_off(sb, type);
2194
2195        err = f2fs_quota_sync(sb, type);
2196        if (err)
2197                goto out_put;
2198
2199        err = dquot_quota_off(sb, type);
2200        if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2201                goto out_put;
2202
2203        inode_lock(inode);
2204        F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2205        f2fs_set_inode_flags(inode);
2206        inode_unlock(inode);
2207        f2fs_mark_inode_dirty_sync(inode, false);
2208out_put:
2209        iput(inode);
2210        return err;
2211}
2212
2213static int f2fs_quota_off(struct super_block *sb, int type)
2214{
2215        struct f2fs_sb_info *sbi = F2FS_SB(sb);
2216        int err;
2217
2218        err = __f2fs_quota_off(sb, type);
2219
2220        /*
2221         * quotactl can shutdown journalled quota, result in inconsistence
2222         * between quota record and fs data by following updates, tag the
2223         * flag to let fsck be aware of it.
2224         */
2225        if (is_journalled_quota(sbi))
2226                set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2227        return err;
2228}
2229
2230void f2fs_quota_off_umount(struct super_block *sb)
2231{
2232        int type;
2233        int err;
2234
2235        for (type = 0; type < MAXQUOTAS; type++) {
2236                err = __f2fs_quota_off(sb, type);
2237                if (err) {
2238                        int ret = dquot_quota_off(sb, type);
2239
2240                        f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2241                                 type, err, ret);
2242                        set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2243                }
2244        }
2245        /*
2246         * In case of checkpoint=disable, we must flush quota blocks.
2247         * This can cause NULL exception for node_inode in end_io, since
2248         * put_super already dropped it.
2249         */
2250        sync_filesystem(sb);
2251}
2252
2253static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2254{
2255        struct quota_info *dqopt = sb_dqopt(sb);
2256        int type;
2257
2258        for (type = 0; type < MAXQUOTAS; type++) {
2259                if (!dqopt->files[type])
2260                        continue;
2261                f2fs_inode_synced(dqopt->files[type]);
2262        }
2263}
2264
2265static int f2fs_dquot_commit(struct dquot *dquot)
2266{
2267        struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2268        int ret;
2269
2270        down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2271        ret = dquot_commit(dquot);
2272        if (ret < 0)
2273                set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2274        up_read(&sbi->quota_sem);
2275        return ret;
2276}
2277
2278static int f2fs_dquot_acquire(struct dquot *dquot)
2279{
2280        struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2281        int ret;
2282
2283        down_read(&sbi->quota_sem);
2284        ret = dquot_acquire(dquot);
2285        if (ret < 0)
2286                set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2287        up_read(&sbi->quota_sem);
2288        return ret;
2289}
2290
2291static int f2fs_dquot_release(struct dquot *dquot)
2292{
2293        struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2294        int ret = dquot_release(dquot);
2295
2296        if (ret < 0)
2297                set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2298        return ret;
2299}
2300
2301static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2302{
2303        struct super_block *sb = dquot->dq_sb;
2304        struct f2fs_sb_info *sbi = F2FS_SB(sb);
2305        int ret = dquot_mark_dquot_dirty(dquot);
2306
2307        /* if we are using journalled quota */
2308        if (is_journalled_quota(sbi))
2309                set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2310
2311        return ret;
2312}
2313
2314static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2315{
2316        struct f2fs_sb_info *sbi = F2FS_SB(sb);
2317        int ret = dquot_commit_info(sb, type);
2318
2319        if (ret < 0)
2320                set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2321        return ret;
2322}
2323
2324static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2325{
2326        *projid = F2FS_I(inode)->i_projid;
2327        return 0;
2328}
2329
2330static const struct dquot_operations f2fs_quota_operations = {
2331        .get_reserved_space = f2fs_get_reserved_space,
2332        .write_dquot    = f2fs_dquot_commit,
2333        .acquire_dquot  = f2fs_dquot_acquire,
2334        .release_dquot  = f2fs_dquot_release,
2335        .mark_dirty     = f2fs_dquot_mark_dquot_dirty,
2336        .write_info     = f2fs_dquot_commit_info,
2337        .alloc_dquot    = dquot_alloc,
2338        .destroy_dquot  = dquot_destroy,
2339        .get_projid     = f2fs_get_projid,
2340        .get_next_id    = dquot_get_next_id,
2341};
2342
2343static const struct quotactl_ops f2fs_quotactl_ops = {
2344        .quota_on       = f2fs_quota_on,
2345        .quota_off      = f2fs_quota_off,
2346        .quota_sync     = f2fs_quota_sync,
2347        .get_state      = dquot_get_state,
2348        .set_info       = dquot_set_dqinfo,
2349        .get_dqblk      = dquot_get_dqblk,
2350        .set_dqblk      = dquot_set_dqblk,
2351        .get_nextdqblk  = dquot_get_next_dqblk,
2352};
2353#else
2354int f2fs_quota_sync(struct super_block *sb, int type)
2355{
2356        return 0;
2357}
2358
2359void f2fs_quota_off_umount(struct super_block *sb)
2360{
2361}
2362#endif
2363
2364static const struct super_operations f2fs_sops = {
2365        .alloc_inode    = f2fs_alloc_inode,
2366        .free_inode     = f2fs_free_inode,
2367        .drop_inode     = f2fs_drop_inode,
2368        .write_inode    = f2fs_write_inode,
2369        .dirty_inode    = f2fs_dirty_inode,
2370        .show_options   = f2fs_show_options,
2371#ifdef CONFIG_QUOTA
2372        .quota_read     = f2fs_quota_read,
2373        .quota_write    = f2fs_quota_write,
2374        .get_dquots     = f2fs_get_dquots,
2375#endif
2376        .evict_inode    = f2fs_evict_inode,
2377        .put_super      = f2fs_put_super,
2378        .sync_fs        = f2fs_sync_fs,
2379        .freeze_fs      = f2fs_freeze,
2380        .unfreeze_fs    = f2fs_unfreeze,
2381        .statfs         = f2fs_statfs,
2382        .remount_fs     = f2fs_remount,
2383};
2384
2385#ifdef CONFIG_FS_ENCRYPTION
2386static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2387{
2388        return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2389                                F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2390                                ctx, len, NULL);
2391}
2392
2393static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2394                                                        void *fs_data)
2395{
2396        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2397
2398        /*
2399         * Encrypting the root directory is not allowed because fsck
2400         * expects lost+found directory to exist and remain unencrypted
2401         * if LOST_FOUND feature is enabled.
2402         *
2403         */
2404        if (f2fs_sb_has_lost_found(sbi) &&
2405                        inode->i_ino == F2FS_ROOT_INO(sbi))
2406                return -EPERM;
2407
2408        return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2409                                F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2410                                ctx, len, fs_data, XATTR_CREATE);
2411}
2412
2413static bool f2fs_dummy_context(struct inode *inode)
2414{
2415        return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode));
2416}
2417
2418static bool f2fs_has_stable_inodes(struct super_block *sb)
2419{
2420        return true;
2421}
2422
2423static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2424                                       int *ino_bits_ret, int *lblk_bits_ret)
2425{
2426        *ino_bits_ret = 8 * sizeof(nid_t);
2427        *lblk_bits_ret = 8 * sizeof(block_t);
2428}
2429
2430static const struct fscrypt_operations f2fs_cryptops = {
2431        .key_prefix             = "f2fs:",
2432        .get_context            = f2fs_get_context,
2433        .set_context            = f2fs_set_context,
2434        .dummy_context          = f2fs_dummy_context,
2435        .empty_dir              = f2fs_empty_dir,
2436        .max_namelen            = F2FS_NAME_LEN,
2437        .has_stable_inodes      = f2fs_has_stable_inodes,
2438        .get_ino_and_lblk_bits  = f2fs_get_ino_and_lblk_bits,
2439};
2440#endif
2441
2442static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2443                u64 ino, u32 generation)
2444{
2445        struct f2fs_sb_info *sbi = F2FS_SB(sb);
2446        struct inode *inode;
2447
2448        if (f2fs_check_nid_range(sbi, ino))
2449                return ERR_PTR(-ESTALE);
2450
2451        /*
2452         * f2fs_iget isn't quite right if the inode is currently unallocated!
2453         * However f2fs_iget currently does appropriate checks to handle stale
2454         * inodes so everything is OK.
2455         */
2456        inode = f2fs_iget(sb, ino);
2457        if (IS_ERR(inode))
2458                return ERR_CAST(inode);
2459        if (unlikely(generation && inode->i_generation != generation)) {
2460                /* we didn't find the right inode.. */
2461                iput(inode);
2462                return ERR_PTR(-ESTALE);
2463        }
2464        return inode;
2465}
2466
2467static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2468                int fh_len, int fh_type)
2469{
2470        return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2471                                    f2fs_nfs_get_inode);
2472}
2473
2474static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2475                int fh_len, int fh_type)
2476{
2477        return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2478                                    f2fs_nfs_get_inode);
2479}
2480
2481static const struct export_operations f2fs_export_ops = {
2482        .fh_to_dentry = f2fs_fh_to_dentry,
2483        .fh_to_parent = f2fs_fh_to_parent,
2484        .get_parent = f2fs_get_parent,
2485};
2486
2487static loff_t max_file_blocks(void)
2488{
2489        loff_t result = 0;
2490        loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2491
2492        /*
2493         * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2494         * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2495         * space in inode.i_addr, it will be more safe to reassign
2496         * result as zero.
2497         */
2498
2499        /* two direct node blocks */
2500        result += (leaf_count * 2);
2501
2502        /* two indirect node blocks */
2503        leaf_count *= NIDS_PER_BLOCK;
2504        result += (leaf_count * 2);
2505
2506        /* one double indirect node block */
2507        leaf_count *= NIDS_PER_BLOCK;
2508        result += leaf_count;
2509
2510        return result;
2511}
2512
2513static int __f2fs_commit_super(struct buffer_head *bh,
2514                        struct f2fs_super_block *super)
2515{
2516        lock_buffer(bh);
2517        if (super)
2518                memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2519        set_buffer_dirty(bh);
2520        unlock_buffer(bh);
2521
2522        /* it's rare case, we can do fua all the time */
2523        return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2524}
2525
2526static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2527                                        struct buffer_head *bh)
2528{
2529        struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2530                                        (bh->b_data + F2FS_SUPER_OFFSET);
2531        struct super_block *sb = sbi->sb;
2532        u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2533        u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2534        u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2535        u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2536        u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2537        u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2538        u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2539        u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2540        u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2541        u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2542        u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2543        u32 segment_count = le32_to_cpu(raw_super->segment_count);
2544        u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2545        u64 main_end_blkaddr = main_blkaddr +
2546                                (segment_count_main << log_blocks_per_seg);
2547        u64 seg_end_blkaddr = segment0_blkaddr +
2548                                (segment_count << log_blocks_per_seg);
2549
2550        if (segment0_blkaddr != cp_blkaddr) {
2551                f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2552                          segment0_blkaddr, cp_blkaddr);
2553                return true;
2554        }
2555
2556        if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2557                                                        sit_blkaddr) {
2558                f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2559                          cp_blkaddr, sit_blkaddr,
2560                          segment_count_ckpt << log_blocks_per_seg);
2561                return true;
2562        }
2563
2564        if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2565                                                        nat_blkaddr) {
2566                f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2567                          sit_blkaddr, nat_blkaddr,
2568                          segment_count_sit << log_blocks_per_seg);
2569                return true;
2570        }
2571
2572        if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2573                                                        ssa_blkaddr) {
2574                f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2575                          nat_blkaddr, ssa_blkaddr,
2576                          segment_count_nat << log_blocks_per_seg);
2577                return true;
2578        }
2579
2580        if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2581                                                        main_blkaddr) {
2582                f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2583                          ssa_blkaddr, main_blkaddr,
2584                          segment_count_ssa << log_blocks_per_seg);
2585                return true;
2586        }
2587
2588        if (main_end_blkaddr > seg_end_blkaddr) {
2589                f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)",
2590                          main_blkaddr,
2591                          segment0_blkaddr +
2592                          (segment_count << log_blocks_per_seg),
2593                          segment_count_main << log_blocks_per_seg);
2594                return true;
2595        } else if (main_end_blkaddr < seg_end_blkaddr) {
2596                int err = 0;
2597                char *res;
2598
2599                /* fix in-memory information all the time */
2600                raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2601                                segment0_blkaddr) >> log_blocks_per_seg);
2602
2603                if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2604                        set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2605                        res = "internally";
2606                } else {
2607                        err = __f2fs_commit_super(bh, NULL);
2608                        res = err ? "failed" : "done";
2609                }
2610                f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%u) block(%u)",
2611                          res, main_blkaddr,
2612                          segment0_blkaddr +
2613                          (segment_count << log_blocks_per_seg),
2614                          segment_count_main << log_blocks_per_seg);
2615                if (err)
2616                        return true;
2617        }
2618        return false;
2619}
2620
2621static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2622                                struct buffer_head *bh)
2623{
2624        block_t segment_count, segs_per_sec, secs_per_zone;
2625        block_t total_sections, blocks_per_seg;
2626        struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2627                                        (bh->b_data + F2FS_SUPER_OFFSET);
2628        unsigned int blocksize;
2629        size_t crc_offset = 0;
2630        __u32 crc = 0;
2631
2632        if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2633                f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2634                          F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2635                return -EINVAL;
2636        }
2637
2638        /* Check checksum_offset and crc in superblock */
2639        if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2640                crc_offset = le32_to_cpu(raw_super->checksum_offset);
2641                if (crc_offset !=
2642                        offsetof(struct f2fs_super_block, crc)) {
2643                        f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2644                                  crc_offset);
2645                        return -EFSCORRUPTED;
2646                }
2647                crc = le32_to_cpu(raw_super->crc);
2648                if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2649                        f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2650                        return -EFSCORRUPTED;
2651                }
2652        }
2653
2654        /* Currently, support only 4KB page cache size */
2655        if (F2FS_BLKSIZE != PAGE_SIZE) {
2656                f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2657                          PAGE_SIZE);
2658                return -EFSCORRUPTED;
2659        }
2660
2661        /* Currently, support only 4KB block size */
2662        blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
2663        if (blocksize != F2FS_BLKSIZE) {
2664                f2fs_info(sbi, "Invalid blocksize (%u), supports only 4KB",
2665                          blocksize);
2666                return -EFSCORRUPTED;
2667        }
2668
2669        /* check log blocks per segment */
2670        if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2671                f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2672                          le32_to_cpu(raw_super->log_blocks_per_seg));
2673                return -EFSCORRUPTED;
2674        }
2675
2676        /* Currently, support 512/1024/2048/4096 bytes sector size */
2677        if (le32_to_cpu(raw_super->log_sectorsize) >
2678                                F2FS_MAX_LOG_SECTOR_SIZE ||
2679                le32_to_cpu(raw_super->log_sectorsize) <
2680                                F2FS_MIN_LOG_SECTOR_SIZE) {
2681                f2fs_info(sbi, "Invalid log sectorsize (%u)",
2682                          le32_to_cpu(raw_super->log_sectorsize));
2683                return -EFSCORRUPTED;
2684        }
2685        if (le32_to_cpu(raw_super->log_sectors_per_block) +
2686                le32_to_cpu(raw_super->log_sectorsize) !=
2687                        F2FS_MAX_LOG_SECTOR_SIZE) {
2688                f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2689                          le32_to_cpu(raw_super->log_sectors_per_block),
2690                          le32_to_cpu(raw_super->log_sectorsize));
2691                return -EFSCORRUPTED;
2692        }
2693
2694        segment_count = le32_to_cpu(raw_super->segment_count);
2695        segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2696        secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2697        total_sections = le32_to_cpu(raw_super->section_count);
2698
2699        /* blocks_per_seg should be 512, given the above check */
2700        blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2701
2702        if (segment_count > F2FS_MAX_SEGMENT ||
2703                                segment_count < F2FS_MIN_SEGMENTS) {
2704                f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2705                return -EFSCORRUPTED;
2706        }
2707
2708        if (total_sections > segment_count ||
2709                        total_sections < F2FS_MIN_SEGMENTS ||
2710                        segs_per_sec > segment_count || !segs_per_sec) {
2711                f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2712                          segment_count, total_sections, segs_per_sec);
2713                return -EFSCORRUPTED;
2714        }
2715
2716        if ((segment_count / segs_per_sec) < total_sections) {
2717                f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2718                          segment_count, segs_per_sec, total_sections);
2719                return -EFSCORRUPTED;
2720        }
2721
2722        if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2723                f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2724                          segment_count, le64_to_cpu(raw_super->block_count));
2725                return -EFSCORRUPTED;
2726        }
2727
2728        if (RDEV(0).path[0]) {
2729                block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
2730                int i = 1;
2731
2732                while (i < MAX_DEVICES && RDEV(i).path[0]) {
2733                        dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
2734                        i++;
2735                }
2736                if (segment_count != dev_seg_count) {
2737                        f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
2738                                        segment_count, dev_seg_count);
2739                        return -EFSCORRUPTED;
2740                }
2741        }
2742
2743        if (secs_per_zone > total_sections || !secs_per_zone) {
2744                f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2745                          secs_per_zone, total_sections);
2746                return -EFSCORRUPTED;
2747        }
2748        if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2749                        raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2750                        (le32_to_cpu(raw_super->extension_count) +
2751                        raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2752                f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2753                          le32_to_cpu(raw_super->extension_count),
2754                          raw_super->hot_ext_count,
2755                          F2FS_MAX_EXTENSION);
2756                return -EFSCORRUPTED;
2757        }
2758
2759        if (le32_to_cpu(raw_super->cp_payload) >
2760                                (blocks_per_seg - F2FS_CP_PACKS)) {
2761                f2fs_info(sbi, "Insane cp_payload (%u > %u)",
2762                          le32_to_cpu(raw_super->cp_payload),
2763                          blocks_per_seg - F2FS_CP_PACKS);
2764                return -EFSCORRUPTED;
2765        }
2766
2767        /* check reserved ino info */
2768        if (le32_to_cpu(raw_super->node_ino) != 1 ||
2769                le32_to_cpu(raw_super->meta_ino) != 2 ||
2770                le32_to_cpu(raw_super->root_ino) != 3) {
2771                f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
2772                          le32_to_cpu(raw_super->node_ino),
2773                          le32_to_cpu(raw_super->meta_ino),
2774                          le32_to_cpu(raw_super->root_ino));
2775                return -EFSCORRUPTED;
2776        }
2777
2778        /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
2779        if (sanity_check_area_boundary(sbi, bh))
2780                return -EFSCORRUPTED;
2781
2782        return 0;
2783}
2784
2785int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
2786{
2787        unsigned int total, fsmeta;
2788        struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
2789        struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2790        unsigned int ovp_segments, reserved_segments;
2791        unsigned int main_segs, blocks_per_seg;
2792        unsigned int sit_segs, nat_segs;
2793        unsigned int sit_bitmap_size, nat_bitmap_size;
2794        unsigned int log_blocks_per_seg;
2795        unsigned int segment_count_main;
2796        unsigned int cp_pack_start_sum, cp_payload;
2797        block_t user_block_count, valid_user_blocks;
2798        block_t avail_node_count, valid_node_count;
2799        int i, j;
2800
2801        total = le32_to_cpu(raw_super->segment_count);
2802        fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
2803        sit_segs = le32_to_cpu(raw_super->segment_count_sit);
2804        fsmeta += sit_segs;
2805        nat_segs = le32_to_cpu(raw_super->segment_count_nat);
2806        fsmeta += nat_segs;
2807        fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
2808        fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
2809
2810        if (unlikely(fsmeta >= total))
2811                return 1;
2812
2813        ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
2814        reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
2815
2816        if (unlikely(fsmeta < F2FS_MIN_SEGMENTS ||
2817                        ovp_segments == 0 || reserved_segments == 0)) {
2818                f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
2819                return 1;
2820        }
2821
2822        user_block_count = le64_to_cpu(ckpt->user_block_count);
2823        segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2824        log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2825        if (!user_block_count || user_block_count >=
2826                        segment_count_main << log_blocks_per_seg) {
2827                f2fs_err(sbi, "Wrong user_block_count: %u",
2828                         user_block_count);
2829                return 1;
2830        }
2831
2832        valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
2833        if (valid_user_blocks > user_block_count) {
2834                f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
2835                         valid_user_blocks, user_block_count);
2836                return 1;
2837        }
2838
2839        valid_node_count = le32_to_cpu(ckpt->valid_node_count);
2840        avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
2841        if (valid_node_count > avail_node_count) {
2842                f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
2843                         valid_node_count, avail_node_count);
2844                return 1;
2845        }
2846
2847        main_segs = le32_to_cpu(raw_super->segment_count_main);
2848        blocks_per_seg = sbi->blocks_per_seg;
2849
2850        for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2851                if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
2852                        le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
2853                        return 1;
2854                for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
2855                        if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2856                                le32_to_cpu(ckpt->cur_node_segno[j])) {
2857                                f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
2858                                         i, j,
2859                                         le32_to_cpu(ckpt->cur_node_segno[i]));
2860                                return 1;
2861                        }
2862                }
2863        }
2864        for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
2865                if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
2866                        le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
2867                        return 1;
2868                for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
2869                        if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
2870                                le32_to_cpu(ckpt->cur_data_segno[j])) {
2871                                f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
2872                                         i, j,
2873                                         le32_to_cpu(ckpt->cur_data_segno[i]));
2874                                return 1;
2875                        }
2876                }
2877        }
2878        for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
2879                for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
2880                        if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
2881                                le32_to_cpu(ckpt->cur_data_segno[j])) {
2882                                f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
2883                                         i, j,
2884                                         le32_to_cpu(ckpt->cur_node_segno[i]));
2885                                return 1;
2886                        }
2887                }
2888        }
2889
2890        sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
2891        nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
2892
2893        if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
2894                nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
2895                f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
2896                         sit_bitmap_size, nat_bitmap_size);
2897                return 1;
2898        }
2899
2900        cp_pack_start_sum = __start_sum_addr(sbi);
2901        cp_payload = __cp_payload(sbi);
2902        if (cp_pack_start_sum < cp_payload + 1 ||
2903                cp_pack_start_sum > blocks_per_seg - 1 -
2904                        NR_CURSEG_TYPE) {
2905                f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
2906                         cp_pack_start_sum);
2907                return 1;
2908        }
2909
2910        if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
2911                le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
2912                f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
2913                          "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
2914                          "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
2915                          le32_to_cpu(ckpt->checksum_offset));
2916                return 1;
2917        }
2918
2919        if (unlikely(f2fs_cp_error(sbi))) {
2920                f2fs_err(sbi, "A bug case: need to run fsck");
2921                return 1;
2922        }
2923        return 0;
2924}
2925
2926static void init_sb_info(struct f2fs_sb_info *sbi)
2927{
2928        struct f2fs_super_block *raw_super = sbi->raw_super;
2929        int i;
2930
2931        sbi->log_sectors_per_block =
2932                le32_to_cpu(raw_super->log_sectors_per_block);
2933        sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
2934        sbi->blocksize = 1 << sbi->log_blocksize;
2935        sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2936        sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
2937        sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2938        sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2939        sbi->total_sections = le32_to_cpu(raw_super->section_count);
2940        sbi->total_node_count =
2941                (le32_to_cpu(raw_super->segment_count_nat) / 2)
2942                        * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
2943        sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
2944        sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
2945        sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
2946        sbi->cur_victim_sec = NULL_SECNO;
2947        sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
2948        sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
2949        sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
2950        sbi->migration_granularity = sbi->segs_per_sec;
2951
2952        sbi->dir_level = DEF_DIR_LEVEL;
2953        sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
2954        sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
2955        sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
2956        sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
2957        sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
2958        sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
2959                                DEF_UMOUNT_DISCARD_TIMEOUT;
2960        clear_sbi_flag(sbi, SBI_NEED_FSCK);
2961
2962        for (i = 0; i < NR_COUNT_TYPE; i++)
2963                atomic_set(&sbi->nr_pages[i], 0);
2964
2965        for (i = 0; i < META; i++)
2966                atomic_set(&sbi->wb_sync_req[i], 0);
2967
2968        INIT_LIST_HEAD(&sbi->s_list);
2969        mutex_init(&sbi->umount_mutex);
2970        init_rwsem(&sbi->io_order_lock);
2971        spin_lock_init(&sbi->cp_lock);
2972
2973        sbi->dirty_device = 0;
2974        spin_lock_init(&sbi->dev_lock);
2975
2976        init_rwsem(&sbi->sb_lock);
2977        init_rwsem(&sbi->pin_sem);
2978}
2979
2980static int init_percpu_info(struct f2fs_sb_info *sbi)
2981{
2982        int err;
2983
2984        err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
2985        if (err)
2986                return err;
2987
2988        err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
2989                                                                GFP_KERNEL);
2990        if (err)
2991                percpu_counter_destroy(&sbi->alloc_valid_block_count);
2992
2993        return err;
2994}
2995
2996#ifdef CONFIG_BLK_DEV_ZONED
2997static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
2998                               void *data)
2999{
3000        struct f2fs_dev_info *dev = data;
3001
3002        if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL)
3003                set_bit(idx, dev->blkz_seq);
3004        return 0;
3005}
3006
3007static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3008{
3009        struct block_device *bdev = FDEV(devi).bdev;
3010        sector_t nr_sectors = bdev->bd_part->nr_sects;
3011        int ret;
3012
3013        if (!f2fs_sb_has_blkzoned(sbi))
3014                return 0;
3015
3016        if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3017                                SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3018                return -EINVAL;
3019        sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3020        if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3021                                __ilog2_u32(sbi->blocks_per_blkz))
3022                return -EINVAL;
3023        sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3024        FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3025                                        sbi->log_blocks_per_blkz;
3026        if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3027                FDEV(devi).nr_blkz++;
3028
3029        FDEV(devi).blkz_seq = f2fs_kzalloc(sbi,
3030                                        BITS_TO_LONGS(FDEV(devi).nr_blkz)
3031                                        * sizeof(unsigned long),
3032                                        GFP_KERNEL);
3033        if (!FDEV(devi).blkz_seq)
3034                return -ENOMEM;
3035
3036        /* Get block zones type */
3037        ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3038                                  &FDEV(devi));
3039        if (ret < 0)
3040                return ret;
3041
3042        return 0;
3043}
3044#endif
3045
3046/*
3047 * Read f2fs raw super block.
3048 * Because we have two copies of super block, so read both of them
3049 * to get the first valid one. If any one of them is broken, we pass
3050 * them recovery flag back to the caller.
3051 */
3052static int read_raw_super_block(struct f2fs_sb_info *sbi,
3053                        struct f2fs_super_block **raw_super,
3054                        int *valid_super_block, int *recovery)
3055{
3056        struct super_block *sb = sbi->sb;
3057        int block;
3058        struct buffer_head *bh;
3059        struct f2fs_super_block *super;
3060        int err = 0;
3061
3062        super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3063        if (!super)
3064                return -ENOMEM;
3065
3066        for (block = 0; block < 2; block++) {
3067                bh = sb_bread(sb, block);
3068                if (!bh) {
3069                        f2fs_err(sbi, "Unable to read %dth superblock",
3070                                 block + 1);
3071                        err = -EIO;
3072                        *recovery = 1;
3073                        continue;
3074                }
3075
3076                /* sanity checking of raw super */
3077                err = sanity_check_raw_super(sbi, bh);
3078                if (err) {
3079                        f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3080                                 block + 1);
3081                        brelse(bh);
3082                        *recovery = 1;
3083                        continue;
3084                }
3085
3086                if (!*raw_super) {
3087                        memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3088                                                        sizeof(*super));
3089                        *valid_super_block = block;
3090                        *raw_super = super;
3091                }
3092                brelse(bh);
3093        }
3094
3095        /* No valid superblock */
3096        if (!*raw_super)
3097                kvfree(super);
3098        else
3099                err = 0;
3100
3101        return err;
3102}
3103
3104int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3105{
3106        struct buffer_head *bh;
3107        __u32 crc = 0;
3108        int err;
3109
3110        if ((recover && f2fs_readonly(sbi->sb)) ||
3111                                bdev_read_only(sbi->sb->s_bdev)) {
3112                set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3113                return -EROFS;
3114        }
3115
3116        /* we should update superblock crc here */
3117        if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3118                crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3119                                offsetof(struct f2fs_super_block, crc));
3120                F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3121        }
3122
3123        /* write back-up superblock first */
3124        bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3125        if (!bh)
3126                return -EIO;
3127        err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3128        brelse(bh);
3129
3130        /* if we are in recovery path, skip writing valid superblock */
3131        if (recover || err)
3132                return err;
3133
3134        /* write current valid superblock */
3135        bh = sb_bread(sbi->sb, sbi->valid_super_block);
3136        if (!bh)
3137                return -EIO;
3138        err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3139        brelse(bh);
3140        return err;
3141}
3142
3143static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3144{
3145        struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3146        unsigned int max_devices = MAX_DEVICES;
3147        int i;
3148
3149        /* Initialize single device information */
3150        if (!RDEV(0).path[0]) {
3151                if (!bdev_is_zoned(sbi->sb->s_bdev))
3152                        return 0;
3153                max_devices = 1;
3154        }
3155
3156        /*
3157         * Initialize multiple devices information, or single
3158         * zoned block device information.
3159         */
3160        sbi->devs = f2fs_kzalloc(sbi,
3161                                 array_size(max_devices,
3162                                            sizeof(struct f2fs_dev_info)),
3163                                 GFP_KERNEL);
3164        if (!sbi->devs)
3165                return -ENOMEM;
3166
3167        for (i = 0; i < max_devices; i++) {
3168
3169                if (i > 0 && !RDEV(i).path[0])
3170                        break;
3171
3172                if (max_devices == 1) {
3173                        /* Single zoned block device mount */
3174                        FDEV(0).bdev =
3175                                blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3176                                        sbi->sb->s_mode, sbi->sb->s_type);
3177                } else {
3178                        /* Multi-device mount */
3179                        memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3180                        FDEV(i).total_segments =
3181                                le32_to_cpu(RDEV(i).total_segments);
3182                        if (i == 0) {
3183                                FDEV(i).start_blk = 0;
3184                                FDEV(i).end_blk = FDEV(i).start_blk +
3185                                    (FDEV(i).total_segments <<
3186                                    sbi->log_blocks_per_seg) - 1 +
3187                                    le32_to_cpu(raw_super->segment0_blkaddr);
3188                        } else {
3189                                FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3190                                FDEV(i).end_blk = FDEV(i).start_blk +
3191                                        (FDEV(i).total_segments <<
3192                                        sbi->log_blocks_per_seg) - 1;
3193                        }
3194                        FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3195                                        sbi->sb->s_mode, sbi->sb->s_type);
3196                }
3197                if (IS_ERR(FDEV(i).bdev))
3198                        return PTR_ERR(FDEV(i).bdev);
3199
3200                /* to release errored devices */
3201                sbi->s_ndevs = i + 1;
3202
3203#ifdef CONFIG_BLK_DEV_ZONED
3204                if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3205                                !f2fs_sb_has_blkzoned(sbi)) {
3206                        f2fs_err(sbi, "Zoned block device feature not enabled\n");
3207                        return -EINVAL;
3208                }
3209                if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3210                        if (init_blkz_info(sbi, i)) {
3211                                f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3212                                return -EINVAL;
3213                        }
3214                        if (max_devices == 1)
3215                                break;
3216                        f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3217                                  i, FDEV(i).path,
3218                                  FDEV(i).total_segments,
3219                                  FDEV(i).start_blk, FDEV(i).end_blk,
3220                                  bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3221                                  "Host-aware" : "Host-managed");
3222                        continue;
3223                }
3224#endif
3225                f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3226                          i, FDEV(i).path,
3227                          FDEV(i).total_segments,
3228                          FDEV(i).start_blk, FDEV(i).end_blk);
3229        }
3230        f2fs_info(sbi,
3231                  "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3232        return 0;
3233}
3234
3235static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3236{
3237#ifdef CONFIG_UNICODE
3238        if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) {
3239                const struct f2fs_sb_encodings *encoding_info;
3240                struct unicode_map *encoding;
3241                __u16 encoding_flags;
3242
3243                if (f2fs_sb_has_encrypt(sbi)) {
3244                        f2fs_err(sbi,
3245                                "Can't mount with encoding and encryption");
3246                        return -EINVAL;
3247                }
3248
3249                if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3250                                          &encoding_flags)) {
3251                        f2fs_err(sbi,
3252                                 "Encoding requested by superblock is unknown");
3253                        return -EINVAL;
3254                }
3255
3256                encoding = utf8_load(encoding_info->version);
3257                if (IS_ERR(encoding)) {
3258                        f2fs_err(sbi,
3259                                 "can't mount with superblock charset: %s-%s "
3260                                 "not supported by the kernel. flags: 0x%x.",
3261                                 encoding_info->name, encoding_info->version,
3262                                 encoding_flags);
3263                        return PTR_ERR(encoding);
3264                }
3265                f2fs_info(sbi, "Using encoding defined by superblock: "
3266                         "%s-%s with flags 0x%hx", encoding_info->name,
3267                         encoding_info->version?:"\b", encoding_flags);
3268
3269                sbi->s_encoding = encoding;
3270                sbi->s_encoding_flags = encoding_flags;
3271                sbi->sb->s_d_op = &f2fs_dentry_ops;
3272        }
3273#else
3274        if (f2fs_sb_has_casefold(sbi)) {
3275                f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3276                return -EINVAL;
3277        }
3278#endif
3279        return 0;
3280}
3281
3282static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3283{
3284        struct f2fs_sm_info *sm_i = SM_I(sbi);
3285
3286        /* adjust parameters according to the volume size */
3287        if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3288                F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3289                sm_i->dcc_info->discard_granularity = 1;
3290                sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3291        }
3292
3293        sbi->readdir_ra = 1;
3294}
3295
3296static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3297{
3298        struct f2fs_sb_info *sbi;
3299        struct f2fs_super_block *raw_super;
3300        struct inode *root;
3301        int err;
3302        bool skip_recovery = false, need_fsck = false;
3303        char *options = NULL;
3304        int recovery, i, valid_super_block;
3305        struct curseg_info *seg_i;
3306        int retry_cnt = 1;
3307
3308try_onemore:
3309        err = -EINVAL;
3310        raw_super = NULL;
3311        valid_super_block = -1;
3312        recovery = 0;
3313
3314        /* allocate memory for f2fs-specific super block info */
3315        sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3316        if (!sbi)
3317                return -ENOMEM;
3318
3319        sbi->sb = sb;
3320
3321        /* Load the checksum driver */
3322        sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3323        if (IS_ERR(sbi->s_chksum_driver)) {
3324                f2fs_err(sbi, "Cannot load crc32 driver.");
3325                err = PTR_ERR(sbi->s_chksum_driver);
3326                sbi->s_chksum_driver = NULL;
3327                goto free_sbi;
3328        }
3329
3330        /* set a block size */
3331        if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3332                f2fs_err(sbi, "unable to set blocksize");
3333                goto free_sbi;
3334        }
3335
3336        err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3337                                                                &recovery);
3338        if (err)
3339                goto free_sbi;
3340
3341        sb->s_fs_info = sbi;
3342        sbi->raw_super = raw_super;
3343
3344        /* precompute checksum seed for metadata */
3345        if (f2fs_sb_has_inode_chksum(sbi))
3346                sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3347                                                sizeof(raw_super->uuid));
3348
3349        /*
3350         * The BLKZONED feature indicates that the drive was formatted with
3351         * zone alignment optimization. This is optional for host-aware
3352         * devices, but mandatory for host-managed zoned block devices.
3353         */
3354#ifndef CONFIG_BLK_DEV_ZONED
3355        if (f2fs_sb_has_blkzoned(sbi)) {
3356                f2fs_err(sbi, "Zoned block device support is not enabled");
3357                err = -EOPNOTSUPP;
3358                goto free_sb_buf;
3359        }
3360#endif
3361        default_options(sbi);
3362        /* parse mount options */
3363        options = kstrdup((const char *)data, GFP_KERNEL);
3364        if (data && !options) {
3365                err = -ENOMEM;
3366                goto free_sb_buf;
3367        }
3368
3369        err = parse_options(sb, options);
3370        if (err)
3371                goto free_options;
3372
3373        sbi->max_file_blocks = max_file_blocks();
3374        sb->s_maxbytes = sbi->max_file_blocks <<
3375                                le32_to_cpu(raw_super->log_blocksize);
3376        sb->s_max_links = F2FS_LINK_MAX;
3377
3378        err = f2fs_setup_casefold(sbi);
3379        if (err)
3380                goto free_options;
3381
3382#ifdef CONFIG_QUOTA
3383        sb->dq_op = &f2fs_quota_operations;
3384        sb->s_qcop = &f2fs_quotactl_ops;
3385        sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3386
3387        if (f2fs_sb_has_quota_ino(sbi)) {
3388                for (i = 0; i < MAXQUOTAS; i++) {
3389                        if (f2fs_qf_ino(sbi->sb, i))
3390                                sbi->nquota_files++;
3391                }
3392        }
3393#endif
3394
3395        sb->s_op = &f2fs_sops;
3396#ifdef CONFIG_FS_ENCRYPTION
3397        sb->s_cop = &f2fs_cryptops;
3398#endif
3399#ifdef CONFIG_FS_VERITY
3400        sb->s_vop = &f2fs_verityops;
3401#endif
3402        sb->s_xattr = f2fs_xattr_handlers;
3403        sb->s_export_op = &f2fs_export_ops;
3404        sb->s_magic = F2FS_SUPER_MAGIC;
3405        sb->s_time_gran = 1;
3406        sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3407                (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3408        memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3409        sb->s_iflags |= SB_I_CGROUPWB;
3410
3411        /* init f2fs-specific super block info */
3412        sbi->valid_super_block = valid_super_block;
3413        init_rwsem(&sbi->gc_lock);
3414        mutex_init(&sbi->writepages);
3415        mutex_init(&sbi->cp_mutex);
3416        mutex_init(&sbi->resize_mutex);
3417        init_rwsem(&sbi->node_write);
3418        init_rwsem(&sbi->node_change);
3419
3420        /* disallow all the data/node/meta page writes */
3421        set_sbi_flag(sbi, SBI_POR_DOING);
3422        spin_lock_init(&sbi->stat_lock);
3423
3424        /* init iostat info */
3425        spin_lock_init(&sbi->iostat_lock);
3426        sbi->iostat_enable = false;
3427
3428        for (i = 0; i < NR_PAGE_TYPE; i++) {
3429                int n = (i == META) ? 1: NR_TEMP_TYPE;
3430                int j;
3431
3432                sbi->write_io[i] =
3433                        f2fs_kmalloc(sbi,
3434                                     array_size(n,
3435                                                sizeof(struct f2fs_bio_info)),
3436                                     GFP_KERNEL);
3437                if (!sbi->write_io[i]) {
3438                        err = -ENOMEM;
3439                        goto free_bio_info;
3440                }
3441
3442                for (j = HOT; j < n; j++) {
3443                        init_rwsem(&sbi->write_io[i][j].io_rwsem);
3444                        sbi->write_io[i][j].sbi = sbi;
3445                        sbi->write_io[i][j].bio = NULL;
3446                        spin_lock_init(&sbi->write_io[i][j].io_lock);
3447                        INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3448                        INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
3449                        init_rwsem(&sbi->write_io[i][j].bio_list_lock);
3450                }
3451        }
3452
3453        init_rwsem(&sbi->cp_rwsem);
3454        init_rwsem(&sbi->quota_sem);
3455        init_waitqueue_head(&sbi->cp_wait);
3456        init_sb_info(sbi);
3457
3458        err = init_percpu_info(sbi);
3459        if (err)
3460                goto free_bio_info;
3461
3462        if (F2FS_IO_ALIGNED(sbi)) {
3463                sbi->write_io_dummy =
3464                        mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3465                if (!sbi->write_io_dummy) {
3466                        err = -ENOMEM;
3467                        goto free_percpu;
3468                }
3469        }
3470
3471        /* init per sbi slab cache */
3472        err = f2fs_init_xattr_caches(sbi);
3473        if (err)
3474                goto free_io_dummy;
3475
3476        /* get an inode for meta space */
3477        sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3478        if (IS_ERR(sbi->meta_inode)) {
3479                f2fs_err(sbi, "Failed to read F2FS meta data inode");
3480                err = PTR_ERR(sbi->meta_inode);
3481                goto free_xattr_cache;
3482        }
3483
3484        err = f2fs_get_valid_checkpoint(sbi);
3485        if (err) {
3486                f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3487                goto free_meta_inode;
3488        }
3489
3490        if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3491                set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3492        if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3493                set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3494                sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3495        }
3496
3497        if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3498                set_sbi_flag(sbi, SBI_NEED_FSCK);
3499
3500        /* Initialize device list */
3501        err = f2fs_scan_devices(sbi);
3502        if (err) {
3503                f2fs_err(sbi, "Failed to find devices");
3504                goto free_devices;
3505        }
3506
3507        err = f2fs_init_post_read_wq(sbi);
3508        if (err) {
3509                f2fs_err(sbi, "Failed to initialize post read workqueue");
3510                goto free_devices;
3511        }
3512
3513        sbi->total_valid_node_count =
3514                                le32_to_cpu(sbi->ckpt->valid_node_count);
3515        percpu_counter_set(&sbi->total_valid_inode_count,
3516                                le32_to_cpu(sbi->ckpt->valid_inode_count));
3517        sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3518        sbi->total_valid_block_count =
3519                                le64_to_cpu(sbi->ckpt->valid_block_count);
3520        sbi->last_valid_block_count = sbi->total_valid_block_count;
3521        sbi->reserved_blocks = 0;
3522        sbi->current_reserved_blocks = 0;
3523        limit_reserve_root(sbi);
3524
3525        for (i = 0; i < NR_INODE_TYPE; i++) {
3526                INIT_LIST_HEAD(&sbi->inode_list[i]);
3527                spin_lock_init(&sbi->inode_lock[i]);
3528        }
3529        mutex_init(&sbi->flush_lock);
3530
3531        f2fs_init_extent_cache_info(sbi);
3532
3533        f2fs_init_ino_entry_info(sbi);
3534
3535        f2fs_init_fsync_node_info(sbi);
3536
3537        /* setup f2fs internal modules */
3538        err = f2fs_build_segment_manager(sbi);
3539        if (err) {
3540                f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3541                         err);
3542                goto free_sm;
3543        }
3544        err = f2fs_build_node_manager(sbi);
3545        if (err) {
3546                f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3547                         err);
3548                goto free_nm;
3549        }
3550
3551        /* For write statistics */
3552        if (sb->s_bdev->bd_part)
3553                sbi->sectors_written_start =
3554                        (u64)part_stat_read(sb->s_bdev->bd_part,
3555                                            sectors[STAT_WRITE]);
3556
3557        /* Read accumulated write IO statistics if exists */
3558        seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3559        if (__exist_node_summaries(sbi))
3560                sbi->kbytes_written =
3561                        le64_to_cpu(seg_i->journal->info.kbytes_written);
3562
3563        f2fs_build_gc_manager(sbi);
3564
3565        err = f2fs_build_stats(sbi);
3566        if (err)
3567                goto free_nm;
3568
3569        /* get an inode for node space */
3570        sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3571        if (IS_ERR(sbi->node_inode)) {
3572                f2fs_err(sbi, "Failed to read node inode");
3573                err = PTR_ERR(sbi->node_inode);
3574                goto free_stats;
3575        }
3576
3577        /* read root inode and dentry */
3578        root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3579        if (IS_ERR(root)) {
3580                f2fs_err(sbi, "Failed to read root inode");
3581                err = PTR_ERR(root);
3582                goto free_node_inode;
3583        }
3584        if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3585                        !root->i_size || !root->i_nlink) {
3586                iput(root);
3587                err = -EINVAL;
3588                goto free_node_inode;
3589        }
3590
3591        sb->s_root = d_make_root(root); /* allocate root dentry */
3592        if (!sb->s_root) {
3593                err = -ENOMEM;
3594                goto free_node_inode;
3595        }
3596
3597        err = f2fs_register_sysfs(sbi);
3598        if (err)
3599                goto free_root_inode;
3600
3601#ifdef CONFIG_QUOTA
3602        /* Enable quota usage during mount */
3603        if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3604                err = f2fs_enable_quotas(sb);
3605                if (err)
3606                        f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3607        }
3608#endif
3609        /* if there are any orphan inodes, free them */
3610        err = f2fs_recover_orphan_inodes(sbi);
3611        if (err)
3612                goto free_meta;
3613
3614        if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3615                goto reset_checkpoint;
3616
3617        /* recover fsynced data */
3618        if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
3619                        !test_opt(sbi, NORECOVERY)) {
3620                /*
3621                 * mount should be failed, when device has readonly mode, and
3622                 * previous checkpoint was not done by clean system shutdown.
3623                 */
3624                if (f2fs_hw_is_readonly(sbi)) {
3625                        if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3626                                err = -EROFS;
3627                                f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3628                                goto free_meta;
3629                        }
3630                        f2fs_info(sbi, "write access unavailable, skipping recovery");
3631                        goto reset_checkpoint;
3632                }
3633
3634                if (need_fsck)
3635                        set_sbi_flag(sbi, SBI_NEED_FSCK);
3636
3637                if (skip_recovery)
3638                        goto reset_checkpoint;
3639
3640                err = f2fs_recover_fsync_data(sbi, false);
3641                if (err < 0) {
3642                        if (err != -ENOMEM)
3643                                skip_recovery = true;
3644                        need_fsck = true;
3645                        f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3646                                 err);
3647                        goto free_meta;
3648                }
3649        } else {
3650                err = f2fs_recover_fsync_data(sbi, true);
3651
3652                if (!f2fs_readonly(sb) && err > 0) {
3653                        err = -EINVAL;
3654                        f2fs_err(sbi, "Need to recover fsync data");
3655                        goto free_meta;
3656                }
3657        }
3658
3659        /*
3660         * If the f2fs is not readonly and fsync data recovery succeeds,
3661         * check zoned block devices' write pointer consistency.
3662         */
3663        if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
3664                err = f2fs_check_write_pointer(sbi);
3665                if (err)
3666                        goto free_meta;
3667        }
3668
3669reset_checkpoint:
3670        /* f2fs_recover_fsync_data() cleared this already */
3671        clear_sbi_flag(sbi, SBI_POR_DOING);
3672
3673        if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3674                err = f2fs_disable_checkpoint(sbi);
3675                if (err)
3676                        goto sync_free_meta;
3677        } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3678                f2fs_enable_checkpoint(sbi);
3679        }
3680
3681        /*
3682         * If filesystem is not mounted as read-only then
3683         * do start the gc_thread.
3684         */
3685        if (F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF && !f2fs_readonly(sb)) {
3686                /* After POR, we can run background GC thread.*/
3687                err = f2fs_start_gc_thread(sbi);
3688                if (err)
3689                        goto sync_free_meta;
3690        }
3691        kvfree(options);
3692
3693        /* recover broken superblock */
3694        if (recovery) {
3695                err = f2fs_commit_super(sbi, true);
3696                f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3697                          sbi->valid_super_block ? 1 : 2, err);
3698        }
3699
3700        f2fs_join_shrinker(sbi);
3701
3702        f2fs_tuning_parameters(sbi);
3703
3704        f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3705                    cur_cp_version(F2FS_CKPT(sbi)));
3706        f2fs_update_time(sbi, CP_TIME);
3707        f2fs_update_time(sbi, REQ_TIME);
3708        clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3709        return 0;
3710
3711sync_free_meta:
3712        /* safe to flush all the data */
3713        sync_filesystem(sbi->sb);
3714        retry_cnt = 0;
3715
3716free_meta:
3717#ifdef CONFIG_QUOTA
3718        f2fs_truncate_quota_inode_pages(sb);
3719        if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
3720                f2fs_quota_off_umount(sbi->sb);
3721#endif
3722        /*
3723         * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
3724         * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
3725         * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
3726         * falls into an infinite loop in f2fs_sync_meta_pages().
3727         */
3728        truncate_inode_pages_final(META_MAPPING(sbi));
3729        /* evict some inodes being cached by GC */
3730        evict_inodes(sb);
3731        f2fs_unregister_sysfs(sbi);
3732free_root_inode:
3733        dput(sb->s_root);
3734        sb->s_root = NULL;
3735free_node_inode:
3736        f2fs_release_ino_entry(sbi, true);
3737        truncate_inode_pages_final(NODE_MAPPING(sbi));
3738        iput(sbi->node_inode);
3739        sbi->node_inode = NULL;
3740free_stats:
3741        f2fs_destroy_stats(sbi);
3742free_nm:
3743        f2fs_destroy_node_manager(sbi);
3744free_sm:
3745        f2fs_destroy_segment_manager(sbi);
3746        f2fs_destroy_post_read_wq(sbi);
3747free_devices:
3748        destroy_device_list(sbi);
3749        kvfree(sbi->ckpt);
3750free_meta_inode:
3751        make_bad_inode(sbi->meta_inode);
3752        iput(sbi->meta_inode);
3753        sbi->meta_inode = NULL;
3754free_xattr_cache:
3755        f2fs_destroy_xattr_caches(sbi);
3756free_io_dummy:
3757        mempool_destroy(sbi->write_io_dummy);
3758free_percpu:
3759        destroy_percpu_info(sbi);
3760free_bio_info:
3761        for (i = 0; i < NR_PAGE_TYPE; i++)
3762                kvfree(sbi->write_io[i]);
3763
3764#ifdef CONFIG_UNICODE
3765        utf8_unload(sbi->s_encoding);
3766#endif
3767free_options:
3768#ifdef CONFIG_QUOTA
3769        for (i = 0; i < MAXQUOTAS; i++)
3770                kvfree(F2FS_OPTION(sbi).s_qf_names[i]);
3771#endif
3772        kvfree(options);
3773free_sb_buf:
3774        kvfree(raw_super);
3775free_sbi:
3776        if (sbi->s_chksum_driver)
3777                crypto_free_shash(sbi->s_chksum_driver);
3778        kvfree(sbi);
3779
3780        /* give only one another chance */
3781        if (retry_cnt > 0 && skip_recovery) {
3782                retry_cnt--;
3783                shrink_dcache_sb(sb);
3784                goto try_onemore;
3785        }
3786        return err;
3787}
3788
3789static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
3790                        const char *dev_name, void *data)
3791{
3792        return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
3793}
3794
3795static void kill_f2fs_super(struct super_block *sb)
3796{
3797        if (sb->s_root) {
3798                struct f2fs_sb_info *sbi = F2FS_SB(sb);
3799
3800                set_sbi_flag(sbi, SBI_IS_CLOSE);
3801                f2fs_stop_gc_thread(sbi);
3802                f2fs_stop_discard_thread(sbi);
3803
3804                if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
3805                                !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3806                        struct cp_control cpc = {
3807                                .reason = CP_UMOUNT,
3808                        };
3809                        f2fs_write_checkpoint(sbi, &cpc);
3810                }
3811
3812                if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
3813                        sb->s_flags &= ~SB_RDONLY;
3814        }
3815        kill_block_super(sb);
3816}
3817
3818static struct file_system_type f2fs_fs_type = {
3819        .owner          = THIS_MODULE,
3820        .name           = "f2fs",
3821        .mount          = f2fs_mount,
3822        .kill_sb        = kill_f2fs_super,
3823        .fs_flags       = FS_REQUIRES_DEV,
3824};
3825MODULE_ALIAS_FS("f2fs");
3826
3827static int __init init_inodecache(void)
3828{
3829        f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
3830                        sizeof(struct f2fs_inode_info), 0,
3831                        SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
3832        if (!f2fs_inode_cachep)
3833                return -ENOMEM;
3834        return 0;
3835}
3836
3837static void destroy_inodecache(void)
3838{
3839        /*
3840         * Make sure all delayed rcu free inodes are flushed before we
3841         * destroy cache.
3842         */
3843        rcu_barrier();
3844        kmem_cache_destroy(f2fs_inode_cachep);
3845}
3846
3847static int __init init_f2fs_fs(void)
3848{
3849        int err;
3850
3851        if (PAGE_SIZE != F2FS_BLKSIZE) {
3852                printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
3853                                PAGE_SIZE, F2FS_BLKSIZE);
3854                return -EINVAL;
3855        }
3856
3857        f2fs_build_trace_ios();
3858
3859        err = init_inodecache();
3860        if (err)
3861                goto fail;
3862        err = f2fs_create_node_manager_caches();
3863        if (err)
3864                goto free_inodecache;
3865        err = f2fs_create_segment_manager_caches();
3866        if (err)
3867                goto free_node_manager_caches;
3868        err = f2fs_create_checkpoint_caches();
3869        if (err)
3870                goto free_segment_manager_caches;
3871        err = f2fs_create_extent_cache();
3872        if (err)
3873                goto free_checkpoint_caches;
3874        err = f2fs_init_sysfs();
3875        if (err)
3876                goto free_extent_cache;
3877        err = register_shrinker(&f2fs_shrinker_info);
3878        if (err)
3879                goto free_sysfs;
3880        err = register_filesystem(&f2fs_fs_type);
3881        if (err)
3882                goto free_shrinker;
3883        f2fs_create_root_stats();
3884        err = f2fs_init_post_read_processing();
3885        if (err)
3886                goto free_root_stats;
3887        err = f2fs_init_bio_entry_cache();
3888        if (err)
3889                goto free_post_read;
3890        err = f2fs_init_bioset();
3891        if (err)
3892                goto free_bio_enrty_cache;
3893        return 0;
3894free_bio_enrty_cache:
3895        f2fs_destroy_bio_entry_cache();
3896free_post_read:
3897        f2fs_destroy_post_read_processing();
3898free_root_stats:
3899        f2fs_destroy_root_stats();
3900        unregister_filesystem(&f2fs_fs_type);
3901free_shrinker:
3902        unregister_shrinker(&f2fs_shrinker_info);
3903free_sysfs:
3904        f2fs_exit_sysfs();
3905free_extent_cache:
3906        f2fs_destroy_extent_cache();
3907free_checkpoint_caches:
3908        f2fs_destroy_checkpoint_caches();
3909free_segment_manager_caches:
3910        f2fs_destroy_segment_manager_caches();
3911free_node_manager_caches:
3912        f2fs_destroy_node_manager_caches();
3913free_inodecache:
3914        destroy_inodecache();
3915fail:
3916        return err;
3917}
3918
3919static void __exit exit_f2fs_fs(void)
3920{
3921        f2fs_destroy_bioset();
3922        f2fs_destroy_bio_entry_cache();
3923        f2fs_destroy_post_read_processing();
3924        f2fs_destroy_root_stats();
3925        unregister_filesystem(&f2fs_fs_type);
3926        unregister_shrinker(&f2fs_shrinker_info);
3927        f2fs_exit_sysfs();
3928        f2fs_destroy_extent_cache();
3929        f2fs_destroy_checkpoint_caches();
3930        f2fs_destroy_segment_manager_caches();
3931        f2fs_destroy_node_manager_caches();
3932        destroy_inodecache();
3933        f2fs_destroy_trace_ios();
3934}
3935
3936module_init(init_f2fs_fs)
3937module_exit(exit_f2fs_fs)
3938
3939MODULE_AUTHOR("Samsung Electronics's Praesto Team");
3940MODULE_DESCRIPTION("Flash Friendly File System");
3941MODULE_LICENSE("GPL");
3942
3943